framework,version,device,op_name,kernel_source,moe_dtype,num_tokens,hidden_size,inter_size,topk,num_experts,moe_tp_size,moe_ep_size,distribution,latency
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,1,power_law_1.01,0.20161983489990237
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,1,power_law_1.01,0.2954892730712891
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,1,power_law_1.01,0.3487744140625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,1,power_law_1.01,0.49053119659423833
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,1,power_law_1.01,0.6219289779663086
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,1,power_law_1.01,0.6217504119873046
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,1,power_law_1.01,0.6159667205810547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,1,power_law_1.01,0.6235372924804687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,1,power_law_1.01,0.6248409652709961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,1,power_law_1.01,0.635544319152832
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,1,power_law_1.01,0.6574694061279297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,1,power_law_1.01,0.6721874999999999
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,1,power_law_1.01,0.7021408081054688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,1,power_law_1.01,0.7948198699951172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,1,power_law_1.01,0.8541004943847657
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,1,power_law_1.01,1.0282643127441407
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,1,power_law_1.01,1.1650252532958985
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,1,power_law_1.01,1.5305970764160155
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,1,power_law_1.01,1.9849459838867187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,1,power_law_1.01,3.097673645019531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,1,power_law_1.01,3.6583770751953124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,1,power_law_1.01,5.407520141601562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,1,power_law_1.01,8.336399536132813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,1,power_law_1.01,12.651589355468749
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,1,power_law_1.01,15.658159179687498
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,1,power_law_1.01,24.116572265625003
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,1,power_law_1.01,34.86220458984375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,1,power_law_1.01,0.1224998378753662
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,1,power_law_1.01,0.16867647171020508
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,1,power_law_1.01,0.19380096435546873
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,1,power_law_1.01,0.2627372741699219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,1,power_law_1.01,0.2955072021484375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,1,power_law_1.01,0.3051807975769043
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,1,power_law_1.01,0.3004748725891113
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,1,power_law_1.01,0.2993164825439453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,1,power_law_1.01,0.3148409652709961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,1,power_law_1.01,0.3085158348083496
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,1,power_law_1.01,0.31716032028198243
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,1,power_law_1.01,0.32691967010498046
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,1,power_law_1.01,0.35939071655273436
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,1,power_law_1.01,0.37227840423583985
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,1,power_law_1.01,0.38495807647705077
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,1,power_law_1.01,0.30493824005126957
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,1,power_law_1.01,0.4109369659423828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,1,power_law_1.01,0.47664001464843747
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,1,power_law_1.01,0.5762271881103516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,1,power_law_1.01,0.7975558471679688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,1,power_law_1.01,0.9223212432861327
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,1,power_law_1.01,1.4994700622558592
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,1,power_law_1.01,1.8938893127441407
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,1,power_law_1.01,3.0147979736328123
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,1,1,power_law_1.01,43.5327880859375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,1,power_law_1.01,3.9019354248046874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,1,power_law_1.01,6.20973876953125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,1,power_law_1.01,8.137693481445313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,1,1,power_law_1.01,11.114173583984375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,1,1,power_law_1.01,0.0548415994644165
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,1,1,power_law_1.01,0.14811200141906739
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,1,1,power_law_1.01,0.14841919898986816
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,1,1,power_law_1.01,0.14922687530517578
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,1,1,power_law_1.01,0.1475648021697998
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,1,1,power_law_1.01,0.1484659194946289
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,1,1,power_law_1.01,0.15148991584777832
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,1,1,power_law_1.01,0.1535923194885254
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,1,1,power_law_1.01,0.08100416183471679
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,1,1,power_law_1.01,0.10552576065063476
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,1,1,power_law_1.01,0.14505151748657225
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,1,1,power_law_1.01,0.15426112174987794
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,1,1,power_law_1.01,0.1629734420776367
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,1,1,power_law_1.01,0.18182144165039063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,1,1,power_law_1.01,0.19072895050048827
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,1,1,power_law_1.01,0.20177663803100584
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,1,1,power_law_1.01,0.22809152603149413
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,1,1,power_law_1.01,0.29417535781860354
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.01,0.3515321731567383
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.01,0.4739494323730469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.01,0.5987494277954102
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.01,0.847004165649414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.01,1.0905490875244142
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.01,1.6789152526855466
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.01,2.3320851135253906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,1,power_law_1.2,0.20626815795898437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,1,power_law_1.2,0.29698303222656247
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,1,power_law_1.2,0.31963712692260743
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,1,power_law_1.2,0.4948473739624023
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,1,power_law_1.2,0.60259521484375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,1,power_law_1.2,0.6328044891357422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,1,power_law_1.2,0.6351980972290039
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,1,power_law_1.2,0.628598403930664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,1,power_law_1.2,0.6408799743652344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,1,power_law_1.2,0.6440857696533203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,1,power_law_1.2,0.6707071685791016
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.01,3.4201394653320314
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,1,power_law_1.2,0.6876319885253906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,1,power_law_1.2,0.7023680114746094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,1,power_law_1.2,0.7951026916503906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,1,power_law_1.2,0.8456582641601562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,1,power_law_1.2,1.0594937896728516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,1,power_law_1.2,1.1665497589111329
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,1,power_law_1.2,1.5921107482910155
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,1,power_law_1.2,1.920233612060547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,1,power_law_1.2,2.7846795654296876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,1,power_law_1.2,4.039641723632813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.01,4.7481515502929685
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,1,power_law_1.2,5.438104858398438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,1,power_law_1.2,8.347235107421875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,1,power_law_1.2,12.36831298828125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,1,power_law_1.2,15.383908691406251
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,1,1,power_law_1.01,5.578313598632812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,1,power_law_1.2,0.12205504417419433
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,1,power_law_1.2,0.16234432220458986
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,1,power_law_1.2,0.17421184539794923
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,1,power_law_1.2,0.265338249206543
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,1,power_law_1.2,0.2824236869812012
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,1,power_law_1.2,0.30271871566772457
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,1,power_law_1.2,0.2963750457763672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,1,power_law_1.2,0.3014463996887207
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,1,power_law_1.2,0.29920320510864257
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,1,power_law_1.2,0.30950336456298827
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,1,power_law_1.2,0.3039993667602539
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,1,power_law_1.2,0.316954231262207
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,1,power_law_1.2,0.31290239334106446
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,1,power_law_1.2,0.36169345855712887
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,1,power_law_1.2,0.38124095916748046
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,1,power_law_1.2,0.3969945526123047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,1,power_law_1.2,0.4075846481323242
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,1,power_law_1.2,0.474218864440918
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,1,power_law_1.2,0.573133430480957
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,1,power_law_1.2,0.7333702087402344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,1,power_law_1.2,0.9790022277832031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,1,power_law_1.2,25.796481933593746
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,1,power_law_1.2,1.4398757934570312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,1,power_law_1.2,1.9760159301757814
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,1,power_law_1.2,2.9689068603515625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,1,power_law_1.2,34.531687011718745
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,1,power_law_1.2,4.101235961914062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,1,power_law_1.2,5.891609497070313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,1,1,power_law_1.2,0.05497087955474853
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,1,1,power_law_1.2,0.07809408187866211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,1,1,power_law_1.2,0.08844096183776855
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,1,1,power_law_1.2,0.14553855895996093
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,1,1,power_law_1.2,0.14318464279174806
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,1,1,power_law_1.2,0.14826687812805175
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,1,1,power_law_1.2,0.14895039558410644
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,1,1,power_law_1.2,0.14762751579284666
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,1,1,power_law_1.2,0.15063936233520508
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,1,1,power_law_1.2,0.15186944007873535
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,1,1,power_law_1.2,0.1536192035675049
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,1,1,power_law_1.2,0.1569337558746338
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,1,1,power_law_1.2,0.16276800155639648
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,1,1,power_law_1.2,0.18197696685791015
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,1,1,power_law_1.2,0.19582399368286132
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,1,1,power_law_1.2,0.2110316848754883
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,1,1,power_law_1.2,43.7113525390625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,1,1,power_law_1.2,0.2988134384155273
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,1,power_law_1.2,8.6534765625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,1,1,power_law_1.2,0.23307455062866209
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,1,1,power_law_1.2,11.27828857421875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.2,0.3587827301025391
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.2,0.4736479949951172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.2,0.5982783889770508
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.2,1.1363775634765625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.2,0.8842809295654297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.2,1.6439494323730468
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.2,3.291665954589844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,1,power_law_1.01,0.3164249610900879
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,1,power_law_1.01,0.45275520324707036
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,1,power_law_1.01,0.5330611038208007
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,1,power_law_1.01,0.805109100341797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,1,power_law_1.01,0.9243110656738281
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,1,power_law_1.01,0.9307481384277343
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,1,power_law_1.01,0.9380928039550781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,1,power_law_1.01,0.9335142517089844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,1,power_law_1.01,0.9441843414306641
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,1,power_law_1.01,0.9862092590332031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,1,power_law_1.01,1.0961561584472657
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,1,power_law_1.01,1.184172134399414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,1,power_law_1.01,1.29842041015625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,1,power_law_1.01,1.419944305419922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,1,power_law_1.01,1.5272473144531251
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,1,power_law_1.01,1.6776666259765627
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,1,power_law_1.01,2.0715283203125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.2,2.2299769592285155
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,1,power_law_1.01,2.726341857910156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,1,power_law_1.01,3.436037902832031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,1,power_law_1.01,5.0148992919921875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,1,power_law_1.01,6.625672607421874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,1,power_law_1.01,10.04264892578125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,1,power_law_1.01,13.400932617187499
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,1,power_law_1.01,19.9521142578125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.2,4.669742126464844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,1,power_law_1.01,27.458437500000002
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,1,1,power_law_1.2,5.862476196289062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,1,power_law_1.01,0.16366912841796874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,1,power_law_1.01,0.2420863914489746
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,1,power_law_1.01,0.2856415939331055
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,1,power_law_1.01,0.4344140625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,1,power_law_1.01,0.4718112182617188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,1,power_law_1.01,0.4727417755126953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,1,power_law_1.01,0.47627326965332034
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,1,power_law_1.01,0.469219856262207
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,1,power_law_1.01,0.4722796630859375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,1,power_law_1.01,0.4825267028808594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,1,power_law_1.01,0.4872159957885742
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,1,power_law_1.01,0.4823884963989258
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,1,power_law_1.01,0.4976908874511719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,1,power_law_1.01,0.5188243103027343
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,1,power_law_1.01,0.5259667205810546
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,1,power_law_1.01,0.5616812896728516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,1,power_law_1.01,0.6178937530517578
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,1,power_law_1.01,0.7678828430175781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,1,power_law_1.01,0.9623910522460937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,1,power_law_1.01,42.4838037109375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,1,power_law_1.01,1.4540870666503907
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,1,power_law_1.01,1.7681248474121094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,1,power_law_1.01,2.82432373046875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,1,power_law_1.01,3.551742858886719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,1,power_law_1.01,5.500986938476562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,1,power_law_1.01,6.770890502929687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,1,power_law_1.01,61.57451171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,1,power_law_1.01,11.389346923828125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,1,power_law_1.01,14.051458740234375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,1,1,power_law_1.01,78.7268505859375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,1,1,power_law_1.01,18.457514648437503
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,1,1,power_law_1.01,0.08036800384521484
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,1,1,power_law_1.01,0.24293184280395508
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,1,1,power_law_1.01,0.24287807464599612
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,1,1,power_law_1.01,0.24434303283691405
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,1,1,power_law_1.01,0.2402252769470215
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,1,1,power_law_1.01,0.241345272064209
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,1,1,power_law_1.01,0.24746944427490236
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,1,1,power_law_1.01,0.25006143569946293
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,1,1,power_law_1.01,0.1327603244781494
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,1,1,power_law_1.01,0.15653887748718262
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,1,1,power_law_1.01,0.23688320159912107
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,1,1,power_law_1.01,0.2475155258178711
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,1,1,power_law_1.01,0.26183616638183593
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,1,1,power_law_1.01,0.29286848068237303
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,1,1,power_law_1.01,0.3059033584594727
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,1,1,power_law_1.01,0.32169982910156253
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,1,1,power_law_1.01,0.37470527648925783
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,1,1,power_law_1.01,0.48708480834960943
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.01,0.7709945678710938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.01,0.600572166442871
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.01,0.9997229003906251
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.01,1.5570419311523438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.01,2.237079620361328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.01,4.0310430908203125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.01,3.3696084594726563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,1,power_law_1.2,0.317193603515625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,1,power_law_1.2,0.4389900970458984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,1,power_law_1.2,0.48871105194091796
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,1,power_law_1.2,0.8099942779541015
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,1,power_law_1.2,0.9100915527343749
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,1,power_law_1.2,0.9422061157226562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,1,power_law_1.2,0.9513177490234375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,1,power_law_1.2,0.9457337951660156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,1,power_law_1.2,0.9653997039794922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,1,power_law_1.2,0.9877606201171876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.01,6.14436279296875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,1,power_law_1.2,1.096024932861328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,1,power_law_1.2,1.2191232299804686
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,1,power_law_1.2,1.291144256591797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,1,power_law_1.2,1.3877395629882812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,1,power_law_1.2,1.5478419494628906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,1,power_law_1.2,1.7287916564941406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,1,power_law_1.2,2.224901123046875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,1,power_law_1.2,2.736661071777344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,1,power_law_1.2,3.7620947265624998
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,1,power_law_1.2,4.858046569824219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,1,power_law_1.2,7.239952392578125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,1,power_law_1.2,10.285111083984376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,1,power_law_1.2,14.452395019531249
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.01,9.228690185546876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,1,power_law_1.2,19.910106201171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,1,1,power_law_1.01,10.657586669921875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,1,power_law_1.2,29.595310058593746
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,1,power_law_1.2,0.16421503067016602
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,1,power_law_1.2,0.23339584350585935
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,1,power_law_1.2,0.26251520156860353
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,1,power_law_1.2,0.43848575592041017
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,1,power_law_1.2,0.4580640029907227
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,1,power_law_1.2,0.4768646240234375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,1,power_law_1.2,0.47917438507080073
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,1,power_law_1.2,0.47437889099121094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,1,power_law_1.2,0.48174270629882815
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,1,power_law_1.2,0.48729217529296875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,1,power_law_1.2,0.4889920043945312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,1,power_law_1.2,0.4931455993652344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,1,power_law_1.2,0.493504638671875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,1,power_law_1.2,0.5215654373168945
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,1,power_law_1.2,0.5353836822509765
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,1,power_law_1.2,0.5643283081054687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,1,power_law_1.2,0.642375717163086
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,1,power_law_1.2,0.7435635375976563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,1,power_law_1.2,0.8664153289794921
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,1,power_law_1.2,1.2423808288574218
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,1,power_law_1.2,1.8038630676269531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,1,power_law_1.2,43.31015625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,1,power_law_1.2,2.6414407348632816
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,1,power_law_1.2,3.700748291015625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,1,power_law_1.2,5.563547973632812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,1,power_law_1.2,7.311259155273437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,1,power_law_1.2,59.9467333984375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,1,1,power_law_1.2,79.64141113281251
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,1,1,power_law_1.2,0.08029696464538574
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,1,1,power_law_1.2,0.12492095947265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,1,1,power_law_1.2,0.14263168334960938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,1,1,power_law_1.2,0.2365785598754883
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,1,1,power_law_1.2,0.23308927536010743
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,1,power_law_1.2,11.510999755859375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,1,1,power_law_1.2,0.24248512268066405
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,1,1,power_law_1.2,0.24331264495849608
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,1,1,power_law_1.2,0.24002815246582032
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,1,1,power_law_1.2,0.24547840118408204
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,1,1,power_law_1.2,0.24634559631347655
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,1,1,power_law_1.2,0.2495590400695801
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,1,1,power_law_1.2,0.25231807708740234
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,1,1,power_law_1.2,0.29559616088867186
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,1,power_law_1.2,14.45078125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,1,1,power_law_1.2,0.3374739074707031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,1,1,power_law_1.2,0.31422336578369137
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,1,1,power_law_1.2,17.199018554687502
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,1,1,power_law_1.2,0.2658950424194336
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,1,1,power_law_1.2,0.3839968109130859
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,1,1,power_law_1.2,0.4891801452636718
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.2,0.5840172958374024
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.2,0.7684326171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.2,1.0760678100585939
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.2,3.0002951049804687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.2,1.5498419189453125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.2,2.0815635681152345
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.2,4.2005178833007815
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,1,power_law_1.01,0.1438950443267822
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,1,power_law_1.01,0.28709375381469726
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,1,power_law_1.01,0.4995942306518555
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,1,power_law_1.01,0.9171993255615234
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,1,power_law_1.01,1.3838400268554687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,1,power_law_1.01,1.793756103515625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,1,power_law_1.01,2.672522888183594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,1,power_law_1.01,2.8497958374023438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,1,power_law_1.01,2.9180895996093748
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.2,6.3034252929687495
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,1,power_law_1.01,3.0105938720703125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,1,power_law_1.01,3.2031552124023435
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,1,power_law_1.01,3.51455810546875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,1,power_law_1.01,3.75182861328125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,1,power_law_1.01,3.9850695800781253
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,1,power_law_1.01,3.89596923828125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,1,power_law_1.01,3.896372985839844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,1,power_law_1.01,4.400653381347657
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,1,power_law_1.01,4.725292663574218
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,1,power_law_1.01,5.593600463867188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,1,power_law_1.01,6.363760375976563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.2,8.157447509765625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,1,power_law_1.01,7.958368530273438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,1,power_law_1.01,9.569471435546875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,1,power_law_1.01,10.884002685546875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,1,power_law_1.01,15.990557861328124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,1,power_law_1.01,18.77275146484375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,1,1,power_law_1.2,10.565975341796875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,1,power_law_1.01,24.6126123046875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,1,power_law_1.01,0.09164480209350587
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,1,power_law_1.01,0.15002623558044434
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,1,power_law_1.01,34.156911621093755
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,1,power_law_1.01,0.4765030288696289
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,1,power_law_1.01,0.9329267120361328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,1,power_law_1.01,0.2667878341674805
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,1,power_law_1.01,0.7111711883544922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,1,1,power_law_1.01,40.995883789062496
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,1,power_law_1.01,1.3464749145507813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,1,power_law_1.01,1.285089874267578
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,1,power_law_1.01,1.4756320190429686
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,1,power_law_1.01,1.3948249816894531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,1,power_law_1.01,1.511986541748047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,1,power_law_1.01,1.539436798095703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,1,power_law_1.01,1.6458924865722657
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,1,power_law_1.01,1.4537266540527345
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,1,power_law_1.01,1.770056915283203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,1,power_law_1.01,1.821210174560547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,1,power_law_1.01,2.061514892578125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,1,power_law_1.01,2.1600607299804686
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,1,power_law_1.01,2.1400743103027344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,1,power_law_1.01,2.5228973388671876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,1,power_law_1.01,3.205234069824219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,1,power_law_1.01,3.2373068237304685
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,1,power_law_1.01,4.263728637695313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,1,power_law_1.01,4.8881298828125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,1,power_law_1.01,6.80579833984375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,1,power_law_1.01,8.276552124023437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,1,power_law_1.01,11.665169677734374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,1,1,power_law_1.01,12.640870361328124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,1,power_law_1.01,0.12368831634521485
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,1,power_law_1.01,0.42617729187011716
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,1,power_law_1.01,0.17475711822509765
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,1,power_law_1.01,0.09021951675415038
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,1,power_law_1.01,0.5431289672851562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,1,power_law_1.01,0.7723725128173828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,1,power_law_1.01,0.8147385406494141
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,1,power_law_1.01,0.8367987060546875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,1,power_law_1.01,0.86949951171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,1,power_law_1.01,0.292412166595459
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,1,power_law_1.01,0.8926201629638673
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,1,power_law_1.01,0.9154124450683593
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,1,power_law_1.01,0.938479995727539
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,1,power_law_1.01,0.9408697509765626
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,1,power_law_1.01,0.9781689453125001
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,1,power_law_1.01,1.0362553405761719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,1,power_law_1.01,1.0527500915527344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,1,power_law_1.01,1.1618093109130858
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.01,1.533402862548828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.01,1.784266815185547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.01,1.2530636596679687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.01,2.037183380126953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.01,2.375274200439453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.01,3.2174816894531246
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.01,4.128602294921874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,1,power_law_1.2,0.14898431777954102
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,1,power_law_1.2,0.2837446403503418
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,1,power_law_1.2,0.49761856079101563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,1,power_law_1.2,0.8462969970703125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,1,power_law_1.2,1.2667558288574219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,1,power_law_1.01,5.799874267578125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,1,power_law_1.2,1.7214361572265624
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,1,power_law_1.01,7.455444946289063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,1,power_law_1.2,2.8959365844726563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,1,power_law_1.2,2.9223168945312503
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,1,power_law_1.2,2.478609924316406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,1,power_law_1.2,3.315309448242188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,1,power_law_1.2,3.3042904663085935
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,1,power_law_1.2,3.291347961425781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,7168,2048,8,256,1,1,power_law_1.01,9.24723876953125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,1,power_law_1.2,3.954698181152344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,1,power_law_1.2,4.068055114746094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,1,power_law_1.2,3.745897521972656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,1,power_law_1.2,4.827673645019532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,1,power_law_1.2,5.172662963867188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,1,power_law_1.2,3.801841430664062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,1,power_law_1.2,5.168569946289063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,1,power_law_1.2,7.199405517578126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,1,power_law_1.2,7.965108642578125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,1,power_law_1.2,9.210421142578124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,1,power_law_1.2,11.224156494140626
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,1,power_law_1.2,16.058626708984374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,1,power_law_1.2,19.492894287109372
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,1,power_law_1.2,0.09235008239746094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,1,power_law_1.2,24.805410156249998
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,1,power_law_1.2,0.265728645324707
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,1,power_law_1.2,0.4416147232055664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,1,power_law_1.2,0.6485478210449218
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,1,power_law_1.2,0.8791980743408203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,1,power_law_1.2,0.1508505630493164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,1,power_law_1.2,1.2698592376708984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,1,power_law_1.2,1.3458822631835938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,1,power_law_1.2,31.892209472656248
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,1,power_law_1.2,1.3656710815429687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,1,power_law_1.2,1.4667570495605469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,1,power_law_1.2,1.469509735107422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,1,power_law_1.2,1.5138803100585938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,1,power_law_1.2,1.581795196533203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,1,power_law_1.2,1.6377587890625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,1,power_law_1.2,2.0266336059570316
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,1,power_law_1.2,2.0140121459960936
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,1,power_law_1.2,2.0479571533203123
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,1,power_law_1.2,2.268305206298828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,1,power_law_1.2,2.465001525878906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,1,power_law_1.2,2.6122207641601562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,1,power_law_1.2,2.972869873046875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,1,power_law_1.2,3.6479244995117184
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,1,power_law_1.2,4.152559509277344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,1,power_law_1.2,4.978668212890625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,1,1,power_law_1.2,43.168671875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,1,power_law_1.2,6.9321704101562505
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,1,power_law_1.2,0.08943488121032714
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,1,power_law_1.2,8.718475952148438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,1,power_law_1.2,0.1235193634033203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,1,power_law_1.2,0.1738444709777832
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,1,power_law_1.2,0.27133440017700194
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,1,power_law_1.2,0.3883500671386719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,1,power_law_1.2,10.68595458984375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,1,power_law_1.2,0.5122367858886718
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,1,power_law_1.2,0.7560665893554688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,1,power_law_1.2,0.7976825714111329
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,1,power_law_1.2,0.8277305603027344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,1,power_law_1.2,0.8469888305664062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,1,power_law_1.2,0.8765894317626953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,1,power_law_1.2,0.9377574157714843
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,1,1,power_law_1.2,14.120882568359374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,1,power_law_1.2,0.9317964935302735
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,1,power_law_1.2,0.9907750701904297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,1,power_law_1.2,1.0723020935058594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,1,power_law_1.2,1.1097248077392579
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,1,power_law_1.2,1.160077438354492
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.2,1.4177778625488282
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,1,power_law_1.2,1.2876287841796876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.2,1.9162162780761718
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.2,2.500767974853516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.2,1.872923583984375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.2,3.146501159667969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.2,4.4260888671875005
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.2,5.8818994140625005
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,1,power_law_1.01,0.03381119966506958
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,1,power_law_1.01,0.08171135902404784
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,1,power_law_1.01,0.1322815990447998
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,1,power_law_1.01,0.1842304039001465
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,1,power_law_1.01,0.1964371109008789
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,1,power_law_1.01,0.19837247848510742
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,1,power_law_1.01,0.11040255546569824
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,1,power_law_1.01,0.21254016876220705
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,1,power_law_1.01,0.21489984512329102
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,1,power_law_1.01,0.23285503387451173
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,1,power_law_1.01,0.23686656951904297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,1,power_law_1.01,0.2386636734008789
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,1,power_law_1.01,0.24689855575561523
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,1,power_law_1.01,0.26561344146728516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,1,power_law_1.01,0.2717535972595215
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,1,power_law_1.01,0.285411205291748
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,1,power_law_1.01,0.31714496612548826
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,1,power_law_1.01,0.4327846527099609
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,1,power_law_1.01,0.3523161697387695
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,1,power_law_1.01,0.5195430374145508
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,1,power_law_1.01,0.6801657867431641
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,1,power_law_1.01,0.8538995361328124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,1,power_law_1.01,1.2055014038085938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,1,power_law_1.01,1.5710899353027343
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,1,power_law_1.01,2.374841613769531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,1,power_law_1.01,3.004239501953125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2048,768,8,128,1,1,power_law_1.01,3.946558837890625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,1,power_law_1.01,0.04191679954528808
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,1,power_law_1.01,0.03371583938598633
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,1,power_law_1.01,0.06446784019470216
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,1,power_law_1.01,0.07929984092712403
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,1,power_law_1.01,0.09610431671142579
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,1,power_law_1.01,0.1344115161895752
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,1,power_law_1.01,0.1387782382965088
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,1,power_law_1.01,0.14057024002075197
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,1,power_law_1.01,0.14419391632080078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,1,power_law_1.01,0.14529088020324707
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,1,power_law_1.01,0.1479263973236084
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,1,power_law_1.01,0.16879615783691407
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,1,power_law_1.01,0.16868736267089843
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,1,power_law_1.01,0.17300800323486326
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,1,power_law_1.01,0.17891199111938477
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,1,power_law_1.01,0.1827008056640625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,1,power_law_1.01,0.1929318428039551
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,1,power_law_1.01,0.2008844757080078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,1,power_law_1.01,0.21726144790649413
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,1,power_law_1.01,0.2521433639526367
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,1,power_law_1.01,0.29340543746948244
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,1,power_law_1.01,0.36758846282958985
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,1,power_law_1.01,0.44040000915527344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,1,power_law_1.01,0.5976089477539063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,1,power_law_1.01,0.7484505462646485
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,1,power_law_1.01,1.065868148803711
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,1,power_law_1.01,1.3577011108398438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,2048,768,8,128,1,1,power_law_1.01,1.6700140380859374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,1,power_law_1.01,0.036595199108123776
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,1,power_law_1.01,0.05490047931671142
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,1,power_law_1.01,0.06057919979095459
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,1,power_law_1.01,0.042064638137817384
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,1,power_law_1.01,0.07842624187469482
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,1,power_law_1.01,0.0822822380065918
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,1,power_law_1.01,0.08316864013671875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,1,power_law_1.01,0.0861888027191162
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,1,power_law_1.01,0.08690112113952637
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,1,power_law_1.01,0.09071040153503418
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,1,power_law_1.01,0.09214079856872559
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,1,power_law_1.01,0.09473919868469238
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,1,power_law_1.01,0.09904191970825196
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,1,power_law_1.01,0.04652607917785644
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,1,power_law_1.01,0.10504128456115722
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,1,power_law_1.01,0.10841343879699708
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,1,power_law_1.01,0.1179270362854004
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,1,power_law_1.01,0.13595264434814452
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,1,power_law_1.01,0.15913920402526854
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,1,power_law_1.01,0.2045702362060547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,1,power_law_1.01,0.2524812889099121
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,1,power_law_1.01,0.3498278427124023
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,1,power_law_1.01,0.4611167907714844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,1,power_law_1.01,0.6707353973388672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,1,power_law_1.01,0.8929030609130859
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,1,power_law_1.01,1.3285784912109375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,1,power_law_1.01,1.7896786499023438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,1,power_law_1.2,0.07707712173461914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,1,power_law_1.2,0.10327296257019043
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2048,768,8,128,1,1,power_law_1.01,2.255818176269531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,1,power_law_1.2,0.12791232109069822
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,1,power_law_1.2,0.1879884719848633
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,1,power_law_1.2,0.19883007049560547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,1,power_law_1.2,0.20735616683959962
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,1,power_law_1.2,0.2163360023498535
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,1,power_law_1.2,0.03721983909606934
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,1,power_law_1.2,0.21869760513305664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,1,power_law_1.2,0.23661567687988283
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,1,power_law_1.2,0.24119039535522463
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,1,power_law_1.2,0.25455039978027344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,1,power_law_1.2,0.26960447311401364
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,1,power_law_1.2,0.2753350448608399
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,1,power_law_1.2,0.28838592529296875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,1,power_law_1.2,0.2427129554748535
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,1,power_law_1.2,0.3273715209960938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,1,power_law_1.2,0.3603334426879883
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,1,power_law_1.2,0.4442598342895508
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,1,power_law_1.2,0.519288330078125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,1,power_law_1.2,0.05428991794586182
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,1,power_law_1.2,0.6922835540771485
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,1,power_law_1.2,0.8655225372314452
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,1,power_law_1.2,1.2050137329101562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,1,power_law_1.2,1.5407557678222656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,1,power_law_1.2,0.03330496072769165
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,1,power_law_1.2,0.0411571216583252
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,1,power_law_1.2,0.06381247997283936
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,1,power_law_1.2,0.07460864067077637
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,1,power_law_1.2,0.0956230354309082
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,1,power_law_1.2,0.13615360260009765
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,1,power_law_1.2,0.1393216037750244
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,1,power_law_1.2,2.2868492126464846
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,1,power_law_1.2,0.14232831954956054
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,1,power_law_1.2,0.14683199882507325
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,1,power_law_1.2,0.14508416175842284
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,1,power_law_1.2,0.1483468818664551
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,1,power_law_1.2,0.17153215408325195
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,1,power_law_1.2,0.17721664428710937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,1,power_law_1.2,0.18379520416259765
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,1,power_law_1.2,0.18678655624389648
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,1,power_law_1.2,0.19441215515136717
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,1,power_law_1.2,0.1673030471801758
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,1,power_law_1.2,0.20236736297607422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,1,power_law_1.2,0.22202880859375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,1,power_law_1.2,0.26143104553222657
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,1,power_law_1.2,0.2940006446838379
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,1,power_law_1.2,0.3711507034301758
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,1,power_law_1.2,0.4458547210693359
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,1,power_law_1.2,0.6028044891357422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,1,power_law_1.2,0.7516320037841797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,1,power_law_1.2,3.1674169921875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,1,power_law_1.2,1.0680281829833984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,1,power_law_1.2,0.03693631887435913
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,1,power_law_1.2,0.04192319869995117
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,1,power_law_1.2,0.04761407852172851
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,1,power_law_1.2,0.052650241851806646
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,1,power_law_1.2,0.05883776187896729
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,1,power_law_1.2,0.0791596794128418
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,1,power_law_1.2,0.08243071556091308
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,1,power_law_1.2,0.08519359588623046
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,1,power_law_1.2,0.08779135704040528
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,1,power_law_1.2,0.08874239921569824
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,1,power_law_1.2,0.09188735961914063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,1,power_law_1.2,0.09507136344909668
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,1,power_law_1.2,0.09756863594055176
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2048,768,8,128,1,1,power_law_1.2,3.835618591308594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,1,power_law_1.2,0.10454463958740234
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,1,power_law_1.2,0.11052991867065429
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,1,power_law_1.2,0.11698240280151366
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,1,power_law_1.2,0.16102975845336914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,1,power_law_1.2,1.3562573242187501
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,1,power_law_1.2,0.1300607967376709
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,1,power_law_1.2,0.1944281578063965
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,1,power_law_1.2,0.26731071472167967
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,1,power_law_1.2,0.33986495971679687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,1,power_law_1.2,0.49863552093505864
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,1,power_law_1.2,0.6627417755126953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,1,power_law_1.2,0.9755686187744141
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,1,power_law_1.2,1.284655303955078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,2048,768,8,128,1,1,power_law_1.2,1.6717919921875002
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,1,power_law_1.01,0.08148351669311524
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,1,power_law_1.01,0.13574784278869628
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,1,power_law_1.01,0.2381305694580078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,1,power_law_1.01,0.337088623046875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,1,power_law_1.01,0.4272281646728516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,1,power_law_1.01,0.64635009765625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,1,power_law_1.01,0.684163818359375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,1,power_law_1.01,0.6957151794433594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,1,power_law_1.01,0.7323987579345703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,1,power_law_1.01,0.7413600158691407
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,1,power_law_1.01,0.8121663665771484
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,1,power_law_1.01,0.8549298858642578
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,1,power_law_1.01,0.8578701019287109
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,1,power_law_1.01,0.8896652984619141
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,1,power_law_1.01,0.9103654479980469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,1,power_law_1.01,0.9301522827148437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,1,power_law_1.01,0.9795558166503906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,1,power_law_1.01,1.1539129638671874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,1,power_law_1.01,1.2382335662841797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,1,power_law_1.2,1.9037753295898436
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,1,power_law_1.01,1.7526495361328123
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,1,power_law_1.01,2.0764863586425784
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,1,power_law_1.01,2.981916198730469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,1,power_law_1.01,3.3974188232421874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,1,power_law_1.01,5.090459594726562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,1,power_law_1.01,6.077583618164063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,1,power_law_1.2,2.5415603637695314
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,1,power_law_1.01,10.047926025390625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,1,power_law_1.01,12.374158935546875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,1,power_law_1.01,0.05703936100006104
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,1,1,power_law_1.01,15.546767578125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,1,power_law_1.01,0.09133824348449707
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,1,power_law_1.01,0.19284223556518554
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,1,power_law_1.01,0.24558975219726564
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,1,power_law_1.01,0.14247679710388184
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,1,power_law_1.01,0.35575103759765625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,1,power_law_1.01,0.3831430435180664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,1,power_law_1.01,0.4085606384277344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,1,power_law_1.01,0.4022348785400391
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,1,power_law_1.01,0.4180166244506836
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,1,power_law_1.01,0.4525478363037109
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,1,power_law_1.01,0.4653657531738281
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,1,power_law_1.01,0.4582175827026367
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,1,power_law_1.01,0.4789606475830078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,1,power_law_1.01,0.4828185653686523
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,1,power_law_1.01,0.5038732910156251
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,1,power_law_1.01,0.5336383819580078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,1,power_law_1.01,0.570136947631836
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,1,power_law_1.01,0.6526764678955078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,1,power_law_1.01,0.3954496002197266
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,1,power_law_1.01,0.7541004943847656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,1,power_law_1.01,0.9959244537353517
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,1,power_law_1.01,1.1564556884765627
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,1,power_law_1.01,1.7499385070800781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,1,power_law_1.01,2.365980224609375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,1,power_law_1.01,3.1629214477539063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,1,power_law_1.01,3.9940026855468753
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,1,1,power_law_1.01,5.190282592773437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,1,power_law_1.01,0.048472962379455566
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,1,power_law_1.01,0.11733695983886719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,1,power_law_1.01,0.14234111785888673
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,1,power_law_1.01,0.08739775657653809
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,1,power_law_1.01,0.21079616546630858
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,1,power_law_1.01,0.22462848663330076
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,1,power_law_1.01,0.23242303848266604
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,1,power_law_1.01,0.06405439853668213
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,1,power_law_1.01,0.2390943908691406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,1,power_law_1.01,0.24090047836303713
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,1,power_law_1.01,0.25114624023437504
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,1,power_law_1.01,0.2554265594482422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,1,power_law_1.01,0.2608844757080078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,1,power_law_1.01,0.2719993591308594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2048,768,8,128,1,1,power_law_1.2,3.1588006591796876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,1,power_law_1.01,0.28510656356811526
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,1,power_law_1.01,0.3159916877746582
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,1,power_law_1.01,0.2939705657958984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,1,power_law_1.01,0.36781505584716795
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.01,0.41668865203857425
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.01,0.5295859146118164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.01,0.6558316802978515
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.01,0.8943743896484374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.01,1.1254086303710937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.01,1.630616912841797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.01,2.154626617431641
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,1,power_law_1.2,0.08303680419921874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,1,power_law_1.2,0.13515968322753907
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,1,power_law_1.2,0.23198656082153318
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,1,power_law_1.2,0.3132102394104004
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,1,power_law_1.2,0.3953254318237305
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,1,power_law_1.2,0.659925765991211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,1,power_law_1.2,0.6924813079833985
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,1,power_law_1.2,0.7223884582519531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,1,power_law_1.2,0.7445279693603515
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.01,3.19891845703125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,1,power_law_1.2,0.755905303955078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,1,power_law_1.2,0.8281337738037109
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,1,power_law_1.2,0.8724761962890625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,1,power_law_1.2,0.8707488250732421
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,1,power_law_1.2,0.9071929931640625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,1,power_law_1.2,0.9305222320556641
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,1,power_law_1.2,0.9523391723632812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,1,power_law_1.2,0.9951026916503907
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,1,power_law_1.2,1.2472083282470703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,1,power_law_1.2,1.339838104248047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,1,power_law_1.2,1.9178387451171877
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.01,4.248965148925781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,1,power_law_1.2,2.081818237304687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,1,power_law_1.2,2.955394592285156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,1,power_law_1.2,3.6631066894531252
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,1,1,power_law_1.01,5.324430541992188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,1,power_law_1.2,4.643963012695313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,1,power_law_1.2,6.532996826171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,1,power_law_1.2,0.05863872051239014
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,1,power_law_1.2,0.09112704277038575
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,1,power_law_1.2,0.13715583801269532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,1,power_law_1.2,0.1800992012023926
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,1,power_law_1.2,0.23491775512695315
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,1,power_law_1.2,0.3730828857421875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,1,power_law_1.2,0.3872019195556641
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,1,power_law_1.2,0.40000961303710936
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,1,power_law_1.2,0.41065216064453125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,1,power_law_1.2,0.4091692733764648
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,1,power_law_1.2,0.4158528137207031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,1,power_law_1.2,0.4622028732299805
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,1,power_law_1.2,0.46347393035888673
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,1,power_law_1.2,0.4807673645019531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,1,power_law_1.2,0.4903168106079102
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,1,power_law_1.2,0.4965439987182617
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,1,power_law_1.2,0.5111590576171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,1,power_law_1.2,0.5431897735595703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,1,power_law_1.2,0.586110725402832
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,1,power_law_1.2,0.6807558441162109
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,1,power_law_1.2,9.7176171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,1,power_law_1.2,0.7685606384277344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,1,power_law_1.2,1.0170188903808595
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,1,power_law_1.2,1.2597376251220704
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,1,power_law_1.2,1.6662527465820312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,1,power_law_1.2,2.205622406005859
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,1,power_law_1.2,12.014920654296876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,1,power_law_1.2,3.1738470458984374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,1,power_law_1.2,0.04847871780395508
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,1,power_law_1.2,0.06320000171661377
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,1,power_law_1.2,0.0847264003753662
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,1,power_law_1.2,0.10867391586303712
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,1,power_law_1.2,0.13377856254577636
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,1,power_law_1.2,0.21420864105224607
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,1,power_law_1.2,0.22717760086059569
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,1,power_law_1.2,0.2357107162475586
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,1,power_law_1.2,0.2434931182861328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,1,power_law_1.2,0.2449126434326172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,1,power_law_1.2,0.2525503921508789
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,1,power_law_1.2,0.2635103988647461
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,1,power_law_1.2,0.269297924041748
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,1,power_law_1.2,0.28434879302978516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,1,power_law_1.2,0.3020857620239258
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,1,power_law_1.2,0.3142560005187988
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,1,1,power_law_1.2,14.764619140625001
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,1,power_law_1.2,0.34814334869384767
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,1,power_law_1.2,0.4222892761230469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,1,power_law_1.2,4.227543640136719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.2,0.49376064300537104
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.2,0.6629830169677734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.2,0.8302528381347656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.2,1.1775430297851561
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,1,1,power_law_1.2,4.873626098632813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.2,1.5382617187499998
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.2,2.2445901489257816
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.2,2.9254656982421876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,1,power_law_1.01,0.1543008041381836
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,1,power_law_1.01,0.31254079818725583
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,1,power_law_1.01,0.5157503890991211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,1,power_law_1.01,0.8350303649902344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,1,power_law_1.01,1.1256556701660156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,1,power_law_1.01,1.8456198120117187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,1,power_law_1.01,2.005359954833984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,1,power_law_1.01,2.1259890747070314
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,1,power_law_1.01,2.1984825134277344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,1,power_law_1.01,2.4033477783203123
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,1,power_law_1.01,2.5466239929199217
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,1,power_law_1.01,2.7819973754882814
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,1,power_law_1.01,2.7710833740234375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.2,4.313641662597656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,1,power_law_1.01,2.942781982421875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,1,power_law_1.01,2.676012268066406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,1,power_law_1.01,3.2262731933593747
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,1,power_law_1.01,3.1612274169921877
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,1,power_law_1.01,4.182600402832032
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,1,power_law_1.01,4.230911865234375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,1,power_law_1.01,5.447762451171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,1,power_law_1.01,5.954887084960937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,1,power_law_1.01,8.950426635742186
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,1,power_law_1.01,10.2354541015625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,1,power_law_1.01,15.137646484375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,1,power_law_1.01,17.681129150390625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.2,5.6991650390625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,1,power_law_1.01,27.271962890625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,1,power_law_1.01,0.11327103614807128
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,1,power_law_1.01,0.1698579216003418
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,1,power_law_1.01,0.28511552810668944
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,1,power_law_1.01,0.4345510482788086
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,1,power_law_1.01,0.586264305114746
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,1,power_law_1.01,0.9392447662353515
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,1,power_law_1.01,1.0068736267089844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,1,power_law_1.01,1.067232666015625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,1,power_law_1.01,1.0938777923583984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,1,power_law_1.01,32.5678955078125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,1,power_law_1.01,1.0810387420654297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,1,power_law_1.01,1.1081868743896484
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,1,power_law_1.01,1.1693926239013672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,1,power_law_1.01,1.3323219299316407
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,1,power_law_1.01,1.3278073120117189
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,1,power_law_1.01,1.3431021118164064
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,1,power_law_1.01,1.4286476135253907
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,1,power_law_1.01,1.4263410949707032
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,1,power_law_1.01,1.7128749084472656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,1,power_law_1.01,1.789482879638672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,1,1,power_law_1.01,39.69964599609375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,1,power_law_1.01,2.072095947265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,1,power_law_1.01,2.328974151611328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,1,power_law_1.01,2.8315557861328124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,1,power_law_1.01,3.7903717041015623
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,1,power_law_1.01,4.5654534912109375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,1,power_law_1.01,5.715215454101562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,1,power_law_1.01,8.917685546875001
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,1,power_law_1.01,10.474509277343751
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,1,1,power_law_1.01,12.7008837890625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,1,1,power_law_1.2,7.1328845214843755
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,1,power_law_1.01,0.08131199836730957
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,1,power_law_1.01,0.3416691207885742
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,1,power_law_1.01,0.11281344413757324
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,1,power_law_1.01,0.5492051315307618
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,1,power_law_1.01,0.5882662582397461
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,1,power_law_1.01,0.6091167831420898
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,1,power_law_1.01,0.635865592956543
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,1,power_law_1.01,0.6276204681396484
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,1,power_law_1.01,0.16965440750122068
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,1,power_law_1.01,0.6620722961425781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,1,power_law_1.01,0.671949462890625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,1,power_law_1.01,0.6799846649169922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,1,power_law_1.01,0.7112064361572266
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,1,power_law_1.01,0.7423942565917969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,1,power_law_1.01,0.7411558532714844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,1,power_law_1.01,0.25531967163085934
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,1,power_law_1.01,0.784221420288086
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.01,0.9769395446777344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,1,power_law_1.01,0.8927648162841797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.01,1.170816650390625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.01,1.3803788757324218
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.01,1.8021708679199218
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.01,2.189008026123047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.01,3.0624359130859373
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.01,3.9129580688476566
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.01,5.747526245117188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,1,power_law_1.2,0.31176639556884767
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,1,power_law_1.2,0.5051372909545899
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,1,power_law_1.2,0.15072575569152832
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,1,power_law_1.2,0.7501177978515625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,1,power_law_1.2,1.0287942504882812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,1,power_law_1.2,1.8056550598144532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,1,power_law_1.2,2.062993927001953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,1,power_law_1.2,1.9217286682128907
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,1,power_law_1.2,2.187190399169922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.01,7.561973266601562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,1,power_law_1.2,2.3247911071777345
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,1,power_law_1.2,2.519525146484375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,1,power_law_1.2,2.9029965209960937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,1,power_law_1.2,2.601566162109375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,1,power_law_1.2,2.66474609375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,1,power_law_1.2,3.18424072265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,1,power_law_1.2,3.0229470825195315
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,1,power_law_1.2,3.2190548706054685
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,1,power_law_1.2,4.370930480957031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,1,power_law_1.2,4.336935119628906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,1,power_law_1.2,5.178898315429687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,1,power_law_1.2,6.798204956054687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,1,power_law_1.2,8.321361694335938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,1,power_law_1.2,10.208701782226562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,2560,8,160,1,1,power_law_1.01,9.534727172851563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,1,power_law_1.2,14.861904296875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,1,power_law_1.2,17.519078369140622
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,1,power_law_1.2,0.11286016464233399
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,1,power_law_1.2,0.1700601577758789
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,1,power_law_1.2,0.27277183532714844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,1,power_law_1.2,0.3916774368286133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,1,power_law_1.2,0.5426841735839844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,1,power_law_1.2,0.9190374755859375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,1,power_law_1.2,0.9773689270019531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,1,power_law_1.2,1.0332313537597657
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,1,power_law_1.2,1.0678502655029296
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,1,power_law_1.2,1.0986335754394532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,1,power_law_1.2,1.1297760009765625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,1,power_law_1.2,1.1501074981689452
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,1,power_law_1.2,1.3295423889160156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,1,power_law_1.2,25.17173095703125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,1,power_law_1.2,1.3860569763183594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,1,power_law_1.2,1.350552978515625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,1,power_law_1.2,1.412574005126953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,1,power_law_1.2,1.4390611267089843
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,1,power_law_1.2,1.5922387695312499
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,1,power_law_1.2,1.7149772644042969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,1,power_law_1.2,2.037193603515625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,1,power_law_1.2,2.160181732177734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,1,power_law_1.2,2.9354565429687502
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,1,power_law_1.2,3.5186727905273436
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,1,power_law_1.2,4.567906494140625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,1,power_law_1.2,5.834312133789062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,1,power_law_1.2,34.84669677734375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,1,power_law_1.2,8.176085815429687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,1,power_law_1.2,0.0814361572265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,1,power_law_1.2,0.1090771198272705
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,1,power_law_1.2,0.16471744537353517
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,1,power_law_1.2,0.23018495559692384
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,1,power_law_1.2,0.31628736495971677
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,1,power_law_1.2,0.5410425567626953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,1,power_law_1.2,0.5808550262451172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,1,power_law_1.2,0.6190457534790039
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,1,power_law_1.2,0.6424813079833984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,1,power_law_1.2,0.6627334594726563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,1,1,power_law_1.2,41.9514697265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,1,power_law_1.2,0.680087661743164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,1,power_law_1.2,10.2705322265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,1,power_law_1.2,0.6889920043945312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,1,power_law_1.2,0.7300434875488281
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,1,power_law_1.2,0.7424825286865234
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,1,power_law_1.2,0.8041324615478516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,1,1,power_law_1.2,13.261323242187501
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,1,power_law_1.2,0.7684793853759766
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,1,power_law_1.2,0.8610527801513672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,1,power_law_1.2,0.9582637023925781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.2,1.3869830322265624
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.2,1.1280799865722657
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.2,1.6694387817382812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.2,2.2870407104492188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.2,2.9470443725585938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.2,4.3526196289062495
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.2,5.74211669921875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,1,power_law_1.01,0.14637503623962403
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,1,power_law_1.01,0.28246976852416994
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,1,power_law_1.01,0.496484489440918
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,1,power_law_1.01,0.909195556640625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,1,power_law_1.01,1.6407128906250001
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,1,power_law_1.01,2.315337524414063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,1,power_law_1.01,2.6937255859375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,1,power_law_1.01,3.5411218261718753
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.2,8.677559204101563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,1,power_law_1.01,3.7518637084960935
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,1,power_law_1.01,4.1140026855468745
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,1,power_law_1.01,4.135963439941406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,1,power_law_1.01,4.492057495117187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,1,power_law_1.01,4.51343994140625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,1,power_law_1.01,5.150670776367187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,1,power_law_1.01,5.154481201171874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,1,power_law_1.01,5.968474731445313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,1,power_law_1.01,5.924547119140625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,1,power_law_1.01,7.178615112304688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,1,power_law_1.01,7.876659545898437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.2,11.4968994140625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,1,power_law_1.01,9.484396972656251
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,1,power_law_1.01,10.482325439453124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,1,power_law_1.01,13.44572998046875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,1,power_law_1.01,17.289739990234374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,1,power_law_1.01,19.969554443359375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,1,power_law_1.01,29.0319580078125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,1,power_law_1.01,0.09599743843078613
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,1,power_law_1.01,0.15248895645141602
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,1,power_law_1.01,0.2691993522644043
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,1,power_law_1.01,0.4773376083374023
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,1,power_law_1.01,0.8443039703369142
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,1,power_law_1.01,1.2527014160156251
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,1,power_law_1.01,35.75659912109375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,1,power_law_1.01,1.4338829040527343
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,1,power_law_1.01,1.8239820861816405
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,1,power_law_1.01,2.0040103149414064
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,1,power_law_1.01,2.196691131591797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,1,power_law_1.01,2.258128662109375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,1,power_law_1.01,2.2158271789550783
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,1,power_law_1.01,2.3900326538085936
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,1,1,power_law_1.01,41.335117187499996
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,1,power_law_1.01,2.3342930603027345
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,1,power_law_1.01,2.4686125183105467
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,1,power_law_1.01,2.351429138183594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,1,power_law_1.01,3.060912780761719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,1,power_law_1.01,3.2351974487304687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,1,power_law_1.01,3.178045349121094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,1,power_law_1.01,3.6853579711914066
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,1,power_law_1.01,4.057522583007812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,1,power_law_1.01,4.8310546875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,1,power_law_1.01,6.505440673828124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,1,power_law_1.01,4.053224182128906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,1,power_law_1.01,7.958798828125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,1,power_law_1.01,9.699038085937499
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,1,power_law_1.01,12.887877197265624
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,1,1,power_law_1.01,13.418601074218751
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,1,power_law_1.01,0.1050432014465332
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,1,power_law_1.01,0.11726464271545409
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,1,power_law_1.01,0.10512831687927246
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,1,power_law_1.01,0.13186495780944824
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,1,power_law_1.01,0.1463199996948242
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,1,power_law_1.01,0.16144832611083984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,1,power_law_1.01,0.1752275276184082
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,1,power_law_1.01,0.18994688034057616
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,1,power_law_1.01,0.2184992027282715
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,1,power_law_1.01,0.10848064422607422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,1,power_law_1.01,0.1074630355834961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,1,power_law_1.01,0.2834560012817383
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,1,power_law_1.01,0.3145708847045898
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,1,power_law_1.01,0.3770476913452149
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,1,power_law_1.01,0.4614745712280273
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,1,power_law_1.01,0.5238201522827148
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,1,power_law_1.01,0.665752944946289
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,1,power_law_1.01,0.9580723571777344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,1,power_law_1.01,1.251403503417969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,1,power_law_1.01,1.8386822509765626
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,1,power_law_1.01,2.4140473937988283
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,1,power_law_1.01,3.5648327636718746
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,1,power_law_1.01,4.760255432128906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,1,power_law_1.2,0.14764800071716308
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,1,power_law_1.2,0.286312313079834
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,1,power_law_1.2,0.4962688064575195
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,1,power_law_1.2,0.9189472198486328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,1,power_law_1.2,1.5235270690917968
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,1,power_law_1.2,2.1677676391601564
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,1,power_law_1.01,7.06470703125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,1,power_law_1.2,2.5820404052734376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,1,power_law_1.2,3.523208312988281
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,1,power_law_1.2,3.8884686279296874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,1,power_law_1.2,3.987365112304688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,1,power_law_1.2,4.103431701660156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,1,power_law_1.2,4.561311950683594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,1,power_law_1.2,4.656144104003906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,1,power_law_1.2,5.099906616210937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,1,power_law_1.2,5.43480712890625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,1,power_law_1.2,5.926112060546875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,1,power_law_1.2,5.844188842773438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,1,power_law_1.2,6.743470458984374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,1,power_law_1.2,7.1622467041015625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,1,power_law_1.2,8.630026245117188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,1,power_law_1.2,9.702637939453124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,1,power_law_1.2,12.013319091796875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,1,power_law_1.2,14.020760498046874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,1,power_law_1.2,16.150311279296876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,1,power_law_1.2,22.071550292968748
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,1,power_law_1.2,0.09766592025756836
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,1,power_law_1.2,0.1534297561645508
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,1,power_law_1.2,0.27194623947143554
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,1,power_law_1.2,0.4822150421142578
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,1,power_law_1.2,0.7844159698486328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,1,power_law_1.2,26.425307617187503
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,1,power_law_1.2,1.1251795196533203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,1,power_law_1.2,1.2918266296386718
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,1,power_law_1.2,1.7708090209960936
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,1,power_law_1.2,2.1063270568847656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,1,power_law_1.2,1.81370361328125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,1,power_law_1.2,2.0066969299316404
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,1,power_law_1.2,37.130727539062505
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,1,power_law_1.2,2.032611236572266
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,1,power_law_1.2,2.2456314086914064
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,1,power_law_1.2,2.5453765869140623
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,1,power_law_1.2,2.236194610595703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,1,power_law_1.2,2.465104675292969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,1,power_law_1.2,3.1253433227539062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,1,power_law_1.2,3.046867980957031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,1,power_law_1.2,3.5894107055664066
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,1,power_law_1.2,3.984105529785156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,1,power_law_1.2,3.838504333496094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,1,1,power_law_1.2,41.083286132812496
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,1,power_law_1.2,4.957729797363281
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,1,power_law_1.2,5.502561645507813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,1,power_law_1.2,6.225405883789063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,1,power_law_1.2,7.555639038085937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,1,power_law_1.2,10.759029541015625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,1,power_law_1.2,0.10587008476257324
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,1,power_law_1.2,12.653925781249999
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,1,power_law_1.2,0.10741632461547852
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,1,power_law_1.2,0.11007871627807617
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,1,1,power_law_1.2,13.81721923828125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,1,power_law_1.2,0.1177900791168213
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,1,power_law_1.2,0.13151616096496582
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,1,power_law_1.2,0.17589120864868163
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,1,power_law_1.2,0.1624799919128418
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,1,power_law_1.2,0.18976896286010742
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,1,power_law_1.2,0.2196966361999512
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,1,power_law_1.2,0.2839251136779785
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,1,power_law_1.2,0.3792825698852539
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,1,power_law_1.2,0.46270721435546874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,1,power_law_1.2,0.5259283065795899
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,1,power_law_1.2,0.10569536209106445
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,1,power_law_1.2,0.14634816169738768
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,1,power_law_1.2,0.31611520767211915
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,1,power_law_1.2,0.670202865600586
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,1,power_law_1.2,0.9609574127197267
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,1,power_law_1.2,1.2484786987304688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,1,power_law_1.2,1.8345817565917968
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,1,power_law_1.2,2.4235494995117186
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,1,power_law_1.2,3.5723147583007817
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,1,power_law_1.2,4.726244506835938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,1,power_law_1.2,7.082448120117188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,1,power_law_1.01,0.10458175659179687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,1,power_law_1.01,0.1653740882873535
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,1,power_law_1.01,0.23707136154174804
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,1,power_law_1.01,0.3122438430786133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,1,power_law_1.01,0.47835838317871093
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,1,power_law_1.01,0.5053107070922851
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,1,power_law_1.01,0.5303507232666016
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,1,power_law_1.01,0.5494361495971679
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,1,power_law_1.01,0.5642233657836914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,1,power_law_1.01,0.5906003189086915
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,1,power_law_1.01,0.6492639923095703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,1,power_law_1.01,0.665389404296875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,1,power_law_1.01,0.6868589019775391
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,1,power_law_1.01,0.694128646850586
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,1,power_law_1.01,0.7033164978027344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,1,power_law_1.01,0.7493740844726562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,1,power_law_1.01,0.8260761260986328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,1,power_law_1.01,0.9224921417236327
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,1,power_law_1.01,1.0443590545654298
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,1,power_law_1.01,1.2261395263671875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,1,power_law_1.01,1.7659373474121094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,1,power_law_1.01,2.3062220764160157
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,1,power_law_1.01,3.1819839477539062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,1,power_law_1.01,5.816743164062499
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,1,power_law_1.01,3.9862106323242186
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,1,power_law_1.01,7.184071655273438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,1,1,power_law_1.01,9.850397949218749
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,1,power_law_1.01,0.04194943904876709
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,1,power_law_1.01,0.11478015899658203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,1,power_law_1.01,0.1660755157470703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,1,power_law_1.01,0.17271999359130857
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,1,power_law_1.01,0.18209856033325195
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,1,power_law_1.01,0.18829952239990236
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,1,power_law_1.01,0.05331071853637696
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,1,power_law_1.01,0.1923731231689453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,1,power_law_1.01,0.1952947235107422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,1,power_law_1.01,0.20180543899536132
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,1,power_law_1.01,0.0682431983947754
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,1,power_law_1.01,0.2067424011230469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,1,power_law_1.01,0.21461503982543945
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,1,power_law_1.01,0.21996032714843752
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,1,power_law_1.01,0.22720703125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,1,power_law_1.01,0.09496447563171387
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,1,power_law_1.01,0.27881919860839843
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,1,power_law_1.01,0.24700544357299803
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,1,power_law_1.01,0.3155308723449707
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,1,power_law_1.01,0.3954982376098633
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,1,power_law_1.01,0.4749779129028321
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,1,power_law_1.01,0.7958617401123047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,1,power_law_1.01,0.6276095962524414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,1,power_law_1.01,1.1473900604248048
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,1,power_law_1.01,1.4882360839843751
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,1,power_law_1.01,2.1860057067871095
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,1,power_law_1.01,2.9061965942382812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,1,power_law_1.2,0.064202880859375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,1,power_law_1.2,0.08664447784423827
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,1,power_law_1.2,0.16220096588134764
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,1,power_law_1.2,0.2187942314147949
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,1,power_law_1.2,0.476445426940918
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,1,power_law_1.2,0.5366713714599609
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,1,power_law_1.2,0.2954150390625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,1,power_law_1.2,0.5057068634033203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,1,power_law_1.2,0.554010238647461
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,1,power_law_1.2,0.6012787246704102
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,1,power_law_1.2,0.6701119995117187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,1,power_law_1.2,0.5698944091796875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,1,power_law_1.2,0.701968002319336
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,1,power_law_1.2,0.7095724487304688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,1,power_law_1.2,0.7480889892578124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2688,1856,6,128,1,1,power_law_1.01,3.6765881347656246
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,1,power_law_1.2,0.7674559783935547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,1,power_law_1.2,0.8378540802001954
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,1,power_law_1.2,0.9218131256103514
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,1,power_law_1.2,0.6841158294677735
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,1,power_law_1.2,1.0881804656982421
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,1,power_law_1.2,1.2817861938476562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,1,power_law_1.2,1.870453186035156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,1,power_law_1.2,2.161630096435547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,1,power_law_1.2,2.9991476440429685
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,1,power_law_1.2,3.959902648925781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,1,power_law_1.2,0.04266816139221191
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,1,power_law_1.2,0.05364927768707275
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,1,power_law_1.2,0.06927999973297119
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,1,power_law_1.2,0.08660351753234863
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,1,power_law_1.2,0.11048128128051757
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,1,power_law_1.2,5.750011596679688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,1,power_law_1.2,0.1657766342163086
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,1,power_law_1.2,0.17644607543945312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,1,power_law_1.2,0.1849407958984375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,1,power_law_1.2,0.1914816093444824
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,1,power_law_1.2,0.19440000534057617
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,1,power_law_1.2,0.2016684722900391
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,1,power_law_1.2,0.20951936721801756
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,1,power_law_1.2,0.21603776931762697
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,1,power_law_1.2,0.2240620803833008
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,1,power_law_1.2,7.1945751953125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,1,power_law_1.2,0.24793472290039062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,1,power_law_1.2,0.2706252861022949
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,1,1,power_law_1.2,9.001400146484375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,1,power_law_1.2,0.3193267250061035
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,1,power_law_1.2,0.37307712554931644
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,1,power_law_1.2,0.2378118324279785
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,1,power_law_1.2,0.48757568359375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,1,power_law_1.2,0.609546890258789
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,1,power_law_1.2,0.8732723236083985
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,1,power_law_1.2,1.1450816345214843
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,1,power_law_1.2,1.6860806274414064
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,1,power_law_1.2,2.271527099609375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,1,power_law_1.2,3.4745611572265624
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,1,power_law_1.01,0.30019712448120117
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,1,power_law_1.01,0.5052191925048828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,1,power_law_1.01,0.9355206298828126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,1,power_law_1.01,1.6291839599609375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,1,power_law_1.01,2.402545928955078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,1,power_law_1.01,3.8877792358398438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,1,power_law_1.01,4.448728332519531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,1,power_law_1.2,4.699336853027344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,1,power_law_1.01,4.4462042236328125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,1,power_law_1.01,4.949095764160156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,1,power_law_1.01,5.2096099853515625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,1,power_law_1.01,5.272947387695313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,1,power_law_1.01,5.888084716796875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2688,1856,6,128,1,1,power_law_1.2,5.820321044921875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,1,power_law_1.01,5.872546997070312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,1,power_law_1.01,7.546460571289063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,1,power_law_1.01,7.971456298828125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,1,power_law_1.01,8.114598388671876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,1,power_law_1.01,9.368004760742187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,1,power_law_1.01,9.10767333984375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,1,power_law_1.01,10.081898803710938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,1,power_law_1.01,11.261708984375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,1,power_law_1.01,13.192717285156249
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,1,power_law_1.01,18.703883056640628
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,1,power_law_1.01,21.8724560546875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,1,power_law_1.01,27.52725341796875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,1,power_law_1.01,34.65693359375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,1,power_law_1.01,0.17475200653076173
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,1,power_law_1.01,0.2803571128845215
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,1,power_law_1.01,0.5127199935913086
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,1,power_law_1.01,50.997348632812496
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,1,power_law_1.01,0.8582943725585939
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,1,power_law_1.01,1.2629011535644532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,1,power_law_1.01,2.0311552429199216
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,1,power_law_1.01,2.2362567138671876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,1,power_law_1.01,2.2980685424804688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,1,power_law_1.01,2.4909925842285157
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,1,power_law_1.01,2.4122239685058595
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,1,power_law_1.01,65.7315380859375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,1,power_law_1.01,2.6773965454101565
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,1,power_law_1.01,2.4964454650878904
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,1,power_law_1.01,2.7702310180664065
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,1,power_law_1.01,2.736596374511719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,1,power_law_1.01,3.078074951171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,1,power_law_1.01,2.6229229736328126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,1,power_law_1.01,3.13150390625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,1,power_law_1.01,3.832749328613281
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,1,power_law_1.01,4.480636901855469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,1,power_law_1.01,4.2826776123046875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,1,power_law_1.01,5.407448120117188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,1,power_law_1.01,6.36455810546875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,1,power_law_1.01,8.211433715820313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,1,power_law_1.01,10.887639160156251
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,1,power_law_1.01,12.73337158203125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,1,power_law_1.01,18.026942138671878
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,1,power_law_1.01,23.683818359375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,1,power_law_1.01,0.16079359054565429
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,1,1,power_law_1.01,25.719821777343746
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,1,power_law_1.01,0.1675372886657715
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,1,power_law_1.01,0.19492544174194334
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,1,power_law_1.01,0.1611609649658203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,1,power_law_1.01,0.21286848068237302
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,1,power_law_1.01,0.2314553642272949
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,1,power_law_1.01,0.24832639694213868
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,1,power_law_1.01,0.26669567108154296
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,1,power_law_1.01,0.30045888900756834
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,1,power_law_1.01,0.1645721626281738
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,1,power_law_1.01,0.17519935607910156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,1,power_law_1.01,0.3725062561035156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,1,power_law_1.01,0.4046822357177734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,1,power_law_1.01,0.5862342453002929
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,1,power_law_1.01,0.47784255981445317
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,1,power_law_1.01,0.6551615905761718
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,1,power_law_1.01,0.838779525756836
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,1,power_law_1.01,1.1969836425781248
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,1,power_law_1.01,1.5724000549316406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,1,power_law_1.01,2.288190155029297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,1,power_law_1.01,3.0020120239257815
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,1,power_law_1.01,4.4391845703125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,1,power_law_1.01,5.903745727539063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,1,power_law_1.01,8.734796142578125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,1,power_law_1.01,11.552464599609376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,1,power_law_1.2,0.3140908813476563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,1,power_law_1.2,0.5050175857543946
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,1,power_law_1.2,1.4677760314941406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,1,power_law_1.2,2.0472154235839843
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,1,power_law_1.2,3.8583981323242185
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,1,power_law_1.2,4.2069982910156245
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,1,power_law_1.2,4.492057495117187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,1,power_law_1.2,4.718649597167969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,1,power_law_1.2,5.041468200683594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,1,power_law_1.2,5.342005615234375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,1,power_law_1.2,5.76662353515625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,1,power_law_1.2,6.191787109375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,1,power_law_1.2,7.125921630859375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,1,power_law_1.2,8.566128540039063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,1,power_law_1.2,7.465995483398437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,1,power_law_1.2,9.656552734375001
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,1,power_law_1.2,8.505162963867187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,1,power_law_1.2,10.67795166015625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,1,power_law_1.2,13.548316650390626
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,1,power_law_1.2,12.45170166015625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,1,power_law_1.2,16.740963134765625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,1,power_law_1.2,21.30478759765625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,1,power_law_1.2,27.773920898437503
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,1,power_law_1.2,35.8617626953125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,1,power_law_1.2,0.17042367935180663
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,1,power_law_1.2,0.28513856887817385
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,1,power_law_1.2,0.5144620895385742
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,1,power_law_1.2,0.7909248352050782
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,1,power_law_1.2,47.469638671874996
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,1,power_law_1.2,1.1297676849365235
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,1,power_law_1.2,2.2471136474609374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,1,power_law_1.2,2.3638758850097656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,1,power_law_1.2,2.419669189453125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,1,power_law_1.2,1.9923738098144532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,1,power_law_1.2,2.552461395263672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,1,power_law_1.2,2.509818267822266
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,1,power_law_1.2,2.759413757324219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,1,power_law_1.2,2.632056884765625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,1,power_law_1.2,2.9772518920898436
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,1,power_law_1.2,2.8823431396484374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,1,power_law_1.2,3.1714990234375002
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,1,power_law_1.2,3.205836791992188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,1,power_law_1.2,3.997237243652344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,1,power_law_1.2,4.910083312988282
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,1,power_law_1.2,4.0663623046875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,1,power_law_1.2,5.505592041015625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,1,power_law_1.2,6.286454467773437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,1,power_law_1.2,8.100816040039062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,1,power_law_1.2,66.99048828125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,1,power_law_1.2,10.221094970703124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,1,power_law_1.2,13.215364990234374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,1,power_law_1.2,18.081599121093753
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,1,power_law_1.2,21.16117919921875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,1,power_law_1.2,0.16281536102294922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,1,power_law_1.2,0.168472957611084
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,1,power_law_1.2,0.17798080444335937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,1,power_law_1.2,0.19645055770874023
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,1,power_law_1.2,0.21442176818847658
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,1,power_law_1.2,0.23192447662353516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,1,power_law_1.2,0.24978431701660156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,1,power_law_1.2,0.2679833602905274
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,1,power_law_1.2,0.15887871742248535
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,1,power_law_1.2,0.30262783050537106
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,1,power_law_1.2,0.16324480056762697
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,1,power_law_1.2,0.3684121704101563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,1,power_law_1.2,0.40776512145996097
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,1,power_law_1.2,0.47674686431884766
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,1,power_law_1.2,0.6567443084716797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,1,power_law_1.2,0.5798880004882813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,1,power_law_1.2,0.8384397125244141
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,1,power_law_1.2,1.2015583801269532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,1,power_law_1.2,1.5630758666992188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,1,power_law_1.2,2.2816685485839843
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,1,power_law_1.2,3.0033728027343747
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,1,power_law_1.2,4.463783569335938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,1,power_law_1.2,5.861004638671875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,1,power_law_1.2,8.71878173828125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,1,power_law_1.2,11.5774462890625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,1,power_law_1.01,0.15004799842834474
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,1,power_law_1.01,0.1577939224243164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,1,power_law_1.01,0.1933113670349121
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,1,power_law_1.01,0.25922048568725586
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,1,power_law_1.01,0.3220441436767578
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,1,power_law_1.01,0.32470336914062503
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,1,power_law_1.01,0.32591102600097654
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,1,power_law_1.01,0.3250201416015625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,1,power_law_1.01,0.32810367584228517
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,1,power_law_1.01,0.3423756790161133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,1,power_law_1.01,0.3792025756835938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,1,power_law_1.01,0.3809996795654297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,1,power_law_1.01,0.39378368377685546
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,1,power_law_1.01,0.4317555236816406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,1,power_law_1.01,0.4707257461547852
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,1,power_law_1.01,0.5430476760864258
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,1,power_law_1.01,0.6196601486206055
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,1,power_law_1.01,0.7672857666015626
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,1,power_law_1.01,0.9414669036865234
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,1,power_law_1.01,1.4109075927734376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,1,power_law_1.01,1.8937478637695313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,1,power_law_1.01,2.6850381469726563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,1,power_law_1.01,3.599896240234375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,1,power_law_1.01,5.310510864257813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,1,power_law_1.01,7.4901904296875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,1,power_law_1.01,0.07059711933135986
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,1,power_law_1.01,0.09638336181640625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,1,power_law_1.01,0.11003135681152343
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,1,power_law_1.01,0.14769536018371582
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,1,power_law_1.01,0.15995712280273439
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,1,power_law_1.01,0.15994367599487305
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,1,power_law_1.01,0.1631711959838867
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,1,power_law_1.01,0.15990976333618162
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,1,power_law_1.01,0.1627359962463379
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,1,power_law_1.01,0.16446144104003907
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,1,power_law_1.01,0.1681273651123047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,1,power_law_1.01,0.16661247253417968
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,1,power_law_1.01,0.17621055603027344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,1,power_law_1.01,0.19713472366333007
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,1,power_law_1.01,0.20689088821411133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,1,power_law_1.01,0.21525312423706056
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,1,power_law_1.01,0.22587007522583008
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,1,power_law_1.01,0.2648678398132324
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,1,power_law_1.01,0.3216582489013672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,1,power_law_1.01,0.41612735748291013
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,1,power_law_1.01,0.4935833740234375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,1,power_law_1.01,15.269105224609376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,1,power_law_1.01,0.6940774536132812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,1,power_law_1.01,0.8799378967285156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,1,power_law_1.01,1.2842707824707031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,1,power_law_1.01,1.9809234619140625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,2,1,power_law_1.01,19.992303466796876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,1,power_law_1.01,2.9242437744140624
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,1,power_law_1.01,4.006925354003906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,2,1,power_law_1.01,5.080873107910156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,2,1,power_law_1.01,0.035971200466156004
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,2,1,power_law_1.01,0.0826790428161621
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,2,1,power_law_1.01,0.08309120178222656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,2,1,power_law_1.01,0.0848851203918457
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,2,1,power_law_1.01,0.05026239871978759
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,2,1,power_law_1.01,0.08428671836853027
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,2,1,power_law_1.01,0.08444992065429688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,2,1,power_law_1.01,0.08584575653076172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,2,1,power_law_1.01,0.08643136024475098
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,2,1,power_law_1.01,0.06172031879425048
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,2,1,power_law_1.01,0.08197759628295899
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,2,1,power_law_1.01,0.08749055862426758
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,2,1,power_law_1.01,0.09044672012329101
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,2,1,power_law_1.01,0.09993215560913085
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,2,1,power_law_1.01,0.10522175788879395
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,2,1,power_law_1.01,0.10884927749633788
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,2,1,power_law_1.01,0.12041728019714355
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,2,1,power_law_1.01,0.15252991676330568
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.01,0.18544000625610352
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.01,0.24432256698608396
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.01,0.3035193634033203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.01,0.42608383178710935
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.01,0.5524934387207031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.01,0.8075872039794921
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.01,1.0675635528564453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,1,power_law_1.2,0.1179097557067871
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,1,power_law_1.2,0.15547967910766602
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,1,power_law_1.2,0.17345855712890626
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,1,power_law_1.2,0.26321407318115236
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,1,power_law_1.2,0.3231628799438476
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,1,power_law_1.2,0.3320851135253906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,1,power_law_1.2,0.3353670501708984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,1,power_law_1.2,0.3321696090698242
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,1,power_law_1.2,0.3413811111450195
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,1,power_law_1.2,0.3589139175415039
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,1,power_law_1.2,0.37649791717529296
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,1,power_law_1.2,0.3810995101928711
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,1,power_law_1.2,0.38727169036865233
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,1,power_law_1.2,0.42156417846679684
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,1,power_law_1.2,0.456165771484375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,1,power_law_1.2,0.5562732696533204
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,1,power_law_1.2,0.6153952026367188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,1,power_law_1.2,0.7674681854248047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,1,power_law_1.2,0.9340415954589844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,1,power_law_1.2,1.3258802795410156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,1,power_law_1.2,1.7996493530273436
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.01,1.6859986877441409
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,1,power_law_1.2,2.6258502197265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.01,2.23453125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,1,power_law_1.2,3.5705938720703125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,1,power_law_1.2,4.900853881835937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,2,1,power_law_1.01,2.756184387207031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,1,power_law_1.2,7.532774658203126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,1,power_law_1.2,0.08716223716735841
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,1,power_law_1.2,0.10005248069763184
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,1,power_law_1.2,0.10438783645629883
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,1,power_law_1.2,0.1506163215637207
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,1,power_law_1.2,0.15536191940307617
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,1,power_law_1.2,0.16171840667724607
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,1,power_law_1.2,0.16250944137573242
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,1,power_law_1.2,0.16292480468749998
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,1,power_law_1.2,0.1657529640197754
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,1,power_law_1.2,0.16610815048217772
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,1,power_law_1.2,0.16879871368408203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,1,power_law_1.2,0.16987775802612304
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,1,power_law_1.2,0.17604480743408205
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,1,power_law_1.2,0.19932416915893553
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,1,power_law_1.2,0.21634048461914063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,1,power_law_1.2,0.2263443183898926
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,1,power_law_1.2,0.2280940818786621
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,1,power_law_1.2,0.26633983612060547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,1,power_law_1.2,0.32551231384277346
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,1,power_law_1.2,0.4043987274169922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,1,power_law_1.2,11.36804931640625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,1,power_law_1.2,0.49318016052246094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,1,power_law_1.2,0.7065555572509765
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,1,power_law_1.2,0.864870376586914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,1,power_law_1.2,1.4343186950683593
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,1,power_law_1.2,14.634544677734374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,1,power_law_1.2,1.8580140686035158
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,2,1,power_law_1.2,18.665740966796875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,2,1,power_law_1.2,0.03739392042160034
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,2,1,power_law_1.2,0.04855807781219482
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,2,1,power_law_1.2,0.05380735874176026
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,2,1,power_law_1.2,0.08329279899597168
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,2,1,power_law_1.2,0.08248064041137695
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,2,1,power_law_1.2,0.084584321975708
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,2,1,power_law_1.2,0.08487615585327149
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,2,1,power_law_1.2,0.08436415672302247
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,2,1,power_law_1.2,0.08548928260803222
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,2,1,power_law_1.2,0.08582592010498047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,2,1,power_law_1.2,0.08659520149230956
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,2,1,power_law_1.2,0.08781824111938477
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,2,1,power_law_1.2,0.0906272029876709
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,2,1,power_law_1.2,0.10053312301635742
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,2,1,power_law_1.2,0.10747776031494141
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,2,1,power_law_1.2,0.11268287658691407
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,1,power_law_1.2,3.0519815063476563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,2,1,power_law_1.2,0.15584959983825683
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,1,power_law_1.2,3.782326965332031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,2,1,power_law_1.2,0.12400768280029298
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.2,0.18725311279296875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.2,0.43833278656005864
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.2,0.2455718421936035
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,2,1,power_law_1.2,5.064993896484375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.2,0.3093292808532715
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.2,0.5603500747680664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.2,0.8172723388671874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.2,1.1098886108398438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,1,power_law_1.01,0.1585472011566162
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,1,power_law_1.01,0.25322303771972654
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,1,power_law_1.01,0.3053356742858887
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,1,power_law_1.01,0.4305900955200196
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,1,power_law_1.01,0.5220678329467774
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,1,power_law_1.01,0.5408243179321289
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,1,power_law_1.01,0.5536902236938477
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,1,power_law_1.01,0.5441791915893555
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,1,power_law_1.01,0.5502265548706055
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,1,power_law_1.01,0.5607609558105469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,1,power_law_1.01,0.587965431213379
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,1,power_law_1.01,0.628424301147461
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,1,power_law_1.01,0.6782694244384766
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,1,power_law_1.01,0.7422431945800781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,1,power_law_1.01,0.7837773132324218
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,1,power_law_1.01,0.8261055755615235
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,1,power_law_1.01,0.9772557067871095
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,1,power_law_1.01,1.3248696899414063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,1,power_law_1.01,1.7186572265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,1,power_law_1.01,2.531396484375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,1,power_law_1.01,3.3008575439453125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.2,1.7191705322265627
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,1,power_law_1.01,4.79400634765625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,1,power_law_1.01,5.9497790527343755
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,1,power_law_1.01,10.118851928710937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.2,2.2636869812011717
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,1,power_law_1.01,12.635386962890625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,1,power_law_1.01,19.983802490234375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,1,power_law_1.01,0.09247103691101075
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,1,power_law_1.01,0.13770432472229005
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,1,power_law_1.01,0.16587711334228517
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,1,power_law_1.01,0.24260160446166995
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,1,power_law_1.01,0.2626726341247559
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,1,power_law_1.01,0.26353151321411133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,1,power_law_1.01,0.2652889633178711
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,1,power_law_1.01,0.26311487197875977
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,1,power_law_1.01,0.2631929588317871
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,1,power_law_1.01,0.2694278335571289
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,1,power_law_1.01,0.2721830368041992
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,1,power_law_1.01,0.26885696411132814
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,1,power_law_1.01,0.27663936614990237
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,1,power_law_1.01,0.2871865653991699
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,1,power_law_1.01,0.29368831634521486
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,1,power_law_1.01,0.305416316986084
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,1,power_law_1.01,0.33805759429931637
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,1,power_law_1.01,0.4073580932617188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,1,power_law_1.01,0.48358207702636713
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,1,power_law_1.01,0.6154611206054688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,2,1,power_law_1.2,2.7439450073242186
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,1,power_law_1.01,0.74429443359375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,1,power_law_1.01,1.1404799652099609
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,1,power_law_1.01,1.6928646850585938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,1,power_law_1.01,27.111345214843748
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,1,power_law_1.01,2.467129669189453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,1,power_law_1.01,3.584286804199219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,2,1,power_law_1.01,33.94390869140625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,1,power_law_1.01,5.028432006835938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,1,power_law_1.01,6.220857543945312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,2,1,power_law_1.01,0.04856959819793701
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,2,1,power_law_1.01,0.13130304336547852
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,2,1,power_law_1.01,0.13413439750671388
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,2,1,power_law_1.01,0.13478848457336426
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,2,1,power_law_1.01,0.1348755168914795
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,2,1,power_law_1.01,0.132424955368042
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,2,1,power_law_1.01,0.1328544044494629
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,2,1,power_law_1.01,8.841780395507813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,2,1,power_law_1.01,0.1360896015167236
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,2,1,power_law_1.01,0.07769216060638427
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,2,1,power_law_1.01,0.0904422378540039
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,2,1,power_law_1.01,0.1380838394165039
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,2,1,power_law_1.01,0.136942081451416
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,2,1,power_law_1.01,0.14187135696411132
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,2,1,power_law_1.01,0.15829119682312012
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,2,1,power_law_1.01,0.15865728378295899
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,2,1,power_law_1.01,0.16648832321166993
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,2,1,power_law_1.01,0.18718847274780273
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,2,1,power_law_1.01,0.23892927169799805
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.01,0.28413375854492184
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.01,0.37700031280517576
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.01,0.4710028839111328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.01,0.6759161376953126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.01,0.8723712158203124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.01,1.28815673828125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.01,1.7212902832031252
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,1,power_law_1.2,0.15688063621520995
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,1,power_law_1.2,0.2400192070007324
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,1,power_law_1.2,0.2745548820495606
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,1,power_law_1.2,0.4319315338134766
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,1,power_law_1.2,0.4944966506958008
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,1,power_law_1.2,0.5339257431030273
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,1,power_law_1.2,0.5465139389038086
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,1,power_law_1.2,0.543815689086914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,1,power_law_1.2,0.5578400039672852
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,1,power_law_1.2,0.5580051040649414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,1,power_law_1.2,0.6023590469360351
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,1,power_law_1.2,0.6460224151611328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,1,power_law_1.2,0.6677222442626953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,1,power_law_1.2,0.7352301025390625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,1,power_law_1.2,0.7786048126220704
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,1,power_law_1.2,0.8329126739501953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,1,power_law_1.2,1.0039385223388673
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,1,power_law_1.2,1.3072288513183594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,1,power_law_1.2,1.724188232421875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,1,power_law_1.2,2.5513381958007812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.01,2.6531219482421875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,1,power_law_1.2,3.2539819335937503
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,1,power_law_1.2,4.485222473144531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,1,power_law_1.2,6.193704223632812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.01,3.8110791015624996
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,1,power_law_1.2,9.415372924804688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,2,1,power_law_1.01,4.6080084228515625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,1,power_law_1.2,13.184132080078125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,1,power_law_1.2,0.09220352172851562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,1,power_law_1.2,0.1337491226196289
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,1,power_law_1.2,0.14702655792236327
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,1,power_law_1.2,0.2407929611206055
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,1,power_law_1.2,0.25102655410766606
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,1,power_law_1.2,0.26328512191772463
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,1,power_law_1.2,0.2646259117126465
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,1,power_law_1.2,0.26167936325073243
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,1,power_law_1.2,0.2672678375244141
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,1,power_law_1.2,0.2698944091796875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,1,power_law_1.2,0.27126527786254884
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,1,power_law_1.2,0.27340864181518554
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,1,power_law_1.2,0.27117055892944336
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,1,power_law_1.2,0.2871436882019043
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,1,power_law_1.2,0.2971641540527344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,1,power_law_1.2,0.3122470474243164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,1,power_law_1.2,0.34783679962158204
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,1,power_law_1.2,0.3992985534667969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,1,power_law_1.2,0.4729510498046875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,1,power_law_1.2,0.6262796783447266
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,1,power_law_1.2,0.7416281890869141
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,1,power_law_1.2,20.111628417968753
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,1,power_law_1.2,1.3390956115722656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,1,power_law_1.2,1.4781478881835937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,1,power_law_1.2,2.6961215209960936
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,1,power_law_1.2,26.181240234375004
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,1,power_law_1.2,3.435221862792969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,1,power_law_1.2,4.9584524536132815
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,2,1,power_law_1.2,0.04795328140258789
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,2,1,power_law_1.2,0.07341248035430908
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,2,1,power_law_1.2,0.08274944305419922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,2,1,power_law_1.2,0.13053248405456544
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,2,1,power_law_1.2,0.1284985637664795
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,2,1,power_law_1.2,0.1335942363739014
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,2,1,power_law_1.2,0.13382847785949709
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,2,1,power_law_1.2,0.13154239654541017
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,2,1,power_law_1.2,0.13479935646057128
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,2,1,power_law_1.2,0.13552960395812988
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,2,1,power_law_1.2,0.13662015914916992
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,2,1,power_law_1.2,0.13837759971618652
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,2,1,power_law_1.2,0.1434489631652832
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,2,1,power_law_1.2,0.15893376350402832
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,2,1,power_law_1.2,34.69577880859375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,2,1,power_law_1.2,0.17573888778686525
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,1,power_law_1.2,6.304909057617188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,2,1,power_law_1.2,0.24133760452270508
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,2,1,power_law_1.2,0.16780992507934572
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,2,1,power_law_1.2,0.1934547233581543
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,2,1,power_law_1.2,8.999786987304688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.2,0.2850828742980957
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.2,0.37990718841552734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.2,0.6826387023925781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.2,0.48525375366210943
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.2,0.882708511352539
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.2,1.4910879516601563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.2,1.9367782592773437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,1,power_law_1.01,0.10184896469116211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,1,power_law_1.01,0.14427647590637208
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,1,power_law_1.01,0.283239688873291
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,1,power_law_1.01,0.4910451126098633
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,1,power_law_1.01,0.7119750213623047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,1,power_law_1.01,0.9152774047851562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,1,power_law_1.01,1.2780108642578125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,1,power_law_1.01,1.3941261291503906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,1,power_law_1.01,1.392367401123047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,1,power_law_1.01,1.4374362182617186
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,1,power_law_1.01,1.5261990356445314
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,1,power_law_1.01,1.6834317016601563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,1,power_law_1.01,1.760198974609375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,1,power_law_1.01,2.065260772705078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.2,2.8096551513671875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,1,power_law_1.01,1.8271334838867186
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,1,power_law_1.01,1.8121682739257814
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,1,power_law_1.01,1.9529466247558596
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,1,power_law_1.01,2.1878515624999997
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,1,power_law_1.01,2.710130615234375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,1,power_law_1.01,2.9241592407226564
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,1,power_law_1.01,3.7799462890625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,1,power_law_1.01,4.2315237426757815
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.2,3.672299499511719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,1,power_law_1.01,5.545565795898438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,1,power_law_1.01,6.856126098632812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,1,power_law_1.01,8.661912231445312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,1,power_law_1.01,13.734266357421877
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,1,power_law_1.01,0.07015488147735596
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,1,power_law_1.01,0.0945683193206787
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,1,power_law_1.01,0.15974143981933592
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,1,power_law_1.01,0.27928768157958983
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,1,power_law_1.01,0.4088524627685547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,1,power_law_1.01,15.418668212890626
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,1,power_law_1.01,0.5351353454589843
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,1,power_law_1.01,0.7540966033935547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,1,power_law_1.01,0.7143341064453125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,1,power_law_1.01,0.7690969848632813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,1,power_law_1.01,0.7758541107177734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,1,power_law_1.01,0.8113613128662109
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,1,power_law_1.01,0.8341625976562501
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,1,power_law_1.01,0.8458246612548829
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,1,power_law_1.01,0.8573766326904296
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,2,1,power_law_1.01,18.796004638671874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,1,power_law_1.01,0.9467488098144532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,1,power_law_1.01,0.9714463806152344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,1,power_law_1.01,0.9918457794189454
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,1,power_law_1.01,1.0256800079345703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,1,power_law_1.01,1.1212191772460938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,1,power_law_1.01,1.268616943359375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,1,power_law_1.01,1.3181414794921875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,1,power_law_1.01,1.7203097534179688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,1,power_law_1.01,1.9514451599121094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,1,power_law_1.01,2.741884765625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,1,power_law_1.01,3.0757107543945312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,1,power_law_1.01,4.233048400878906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,1,power_law_1.01,5.921849365234375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,2,1,power_law_1.01,6.470690307617187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,2,1,power_law_1.2,4.792179260253906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,1,power_law_1.01,0.07161215782165528
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,1,power_law_1.01,0.2460531234741211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,1,power_law_1.01,0.09145983695983886
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,1,power_law_1.01,0.3088019180297851
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,1,power_law_1.01,0.42582656860351564
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,1,power_law_1.01,0.448408317565918
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,1,power_law_1.01,0.46111167907714845
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,1,power_law_1.01,0.4660147094726563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,1,power_law_1.01,0.49057792663574223
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,1,power_law_1.01,0.5062905502319336
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,1,power_law_1.01,0.5174540710449219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,1,power_law_1.01,0.11932352066040039
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,1,power_law_1.01,0.5291763305664062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,1,power_law_1.01,0.5536441421508789
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,1,power_law_1.01,0.5609721755981445
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,1,power_law_1.01,0.1741747283935547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,1,power_law_1.01,0.5930342483520508
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,1,power_law_1.01,0.6531897735595703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.01,0.8457868957519532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.01,0.9801254272460938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.01,0.7180339050292969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.01,1.2540006256103515
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.01,1.536944580078125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.01,2.117401580810547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.01,2.8489825439453123
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,1,power_law_1.2,0.1009228801727295
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,1,power_law_1.2,0.1411520004272461
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,1,power_law_1.2,0.2808371162414551
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,1,power_law_1.2,0.4464851379394531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,1,power_law_1.2,0.6515961456298828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,1,power_law_1.2,0.864734115600586
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,1,power_law_1.2,1.2546956634521484
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,1,power_law_1.01,4.082237548828125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,1,power_law_1.2,1.3145964050292969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,1,power_law_1.01,5.222714233398437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,1,power_law_1.2,1.422024383544922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,1,power_law_1.2,1.422216339111328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,1,power_law_1.2,1.5834002685546875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,1,power_law_1.2,1.4873362731933593
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,1,power_law_1.2,1.8991404724121093
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,1,power_law_1.2,1.9045336914062498
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,1,power_law_1.2,1.8659519958496094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,1,power_law_1.2,1.9602758789062498
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,7168,2048,8,256,2,1,power_law_1.01,6.514987792968751
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,1,power_law_1.2,1.960865936279297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,1,power_law_1.2,2.553014373779297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,1,power_law_1.2,2.8370181274414064
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,1,power_law_1.2,3.004913940429687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,1,power_law_1.2,3.4807525634765626
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,1,power_law_1.2,4.6530957031249995
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,1,power_law_1.2,5.716181030273438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,1,power_law_1.2,6.8943505859375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,1,power_law_1.2,8.887540283203126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,1,power_law_1.2,0.06928256034851074
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,1,power_law_1.2,13.08470458984375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,1,power_law_1.2,0.09396736145019531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,1,power_law_1.2,16.317359619140625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,1,power_law_1.2,0.25839231491088865
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,1,power_law_1.2,0.3735948944091797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,1,power_law_1.2,0.5112934494018555
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,1,power_law_1.2,0.7068819427490235
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,1,power_law_1.2,0.7320396423339843
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,1,power_law_1.2,0.7571942138671874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,1,power_law_1.2,0.1562816047668457
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,1,power_law_1.2,0.7774732971191407
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,1,power_law_1.2,0.8022688293457032
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,1,power_law_1.2,0.8301900482177734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,1,power_law_1.2,0.8447526550292969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,1,power_law_1.2,0.8798445129394532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,1,power_law_1.2,0.9715052795410155
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,1,power_law_1.2,0.979056625366211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,2,1,power_law_1.2,18.940694580078123
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,1,power_law_1.2,1.038507537841797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,1,power_law_1.2,1.1035558319091796
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,1,power_law_1.2,1.1373817443847656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,1,power_law_1.2,1.3230668640136718
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,1,power_law_1.2,1.5702943420410156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,1,power_law_1.2,1.7518348693847656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,1,power_law_1.2,2.0099066162109374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,1,power_law_1.2,2.60943359375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,1,power_law_1.2,3.4675277709960937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,1,power_law_1.2,4.439863586425782
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,1,power_law_1.2,0.07224768161773681
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,1,power_law_1.2,0.09075967788696289
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,1,power_law_1.2,0.11732288360595704
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,1,power_law_1.2,5.416538696289062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,1,power_law_1.2,0.16362367630004884
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,1,power_law_1.2,0.22672767639160157
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,2,1,power_law_1.2,6.400784301757812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,1,power_law_1.2,0.29185983657836917
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,1,power_law_1.2,0.41863552093505857
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,1,power_law_1.2,0.4395142364501953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,1,power_law_1.2,0.4545977783203125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,1,power_law_1.2,0.4723104095458984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,1,power_law_1.2,0.49013118743896483
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,1,power_law_1.2,0.5079935836791992
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,1,power_law_1.2,0.5242297744750977
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,1,power_law_1.2,0.5501030349731446
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,1,power_law_1.2,0.574576644897461
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,1,power_law_1.2,0.5993555068969727
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.2,0.8345305633544922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,1,power_law_1.2,0.6386726379394532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.2,1.0265401458740233
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.2,1.228728942871094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,1,power_law_1.2,0.7343007659912109
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.2,1.6679200744628908
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.2,2.1870693969726562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.2,3.1002874755859375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,1,power_law_1.01,0.041899518966674806
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,1,power_law_1.01,0.05458687782287598
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,1,power_law_1.01,0.06835008144378663
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,1,power_law_1.01,0.08095871925354003
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,1,power_law_1.01,0.1165395164489746
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,1,power_law_1.01,0.12244928359985352
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,1,power_law_1.01,0.12469440460205079
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,1,power_law_1.01,0.12743231773376465
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,1,power_law_1.01,0.12769344329833984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.2,4.145028991699219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,1,power_law_1.01,0.13803775787353517
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,1,power_law_1.01,0.1457868766784668
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,1,power_law_1.01,0.1506611156463623
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,1,power_law_1.01,0.15129088401794433
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,1,power_law_1.01,0.15785280227661133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,1,power_law_1.01,0.15759103775024413
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,1,power_law_1.01,0.1656671905517578
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,1,power_law_1.01,0.19746368408203124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,1,power_law_1.01,0.21585472106933593
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,1,power_law_1.01,0.2610220718383789
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,1,power_law_1.01,0.30831552505493165
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,1,power_law_1.01,0.40834495544433597
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,1,power_law_1.01,0.5037753677368164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,1,power_law_1.01,0.7068262481689453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,1,power_law_1.01,0.8962310028076171
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,1,power_law_1.01,1.2923301696777343
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,1,power_law_1.01,1.664934387207031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,1,power_law_1.01,0.032189440727233884
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,1,power_law_1.01,0.03508415937423706
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2048,768,8,128,2,1,power_law_1.01,2.0643212890625002
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,1,power_law_1.01,0.04624832153320312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,1,power_law_1.01,0.0724896001815796
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,1,power_law_1.01,0.09221632003784179
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,1,power_law_1.01,0.059194240570068356
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,1,power_law_1.01,0.09859840393066406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,1,power_law_1.01,0.10076095581054687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,1,power_law_1.01,0.10226688385009766
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,1,power_law_1.01,0.10301952362060547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,1,power_law_1.01,0.10376383781433104
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,1,power_law_1.01,0.11862336158752443
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,1,power_law_1.01,0.11985792160034178
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,1,power_law_1.01,0.12172672271728516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,1,power_law_1.01,0.12416768074035645
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,1,power_law_1.01,0.12655936241149904
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,1,power_law_1.01,0.13313023567199705
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,1,power_law_1.01,0.13940799713134766
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,1,power_law_1.01,0.1541113567352295
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,1,power_law_1.01,0.17746047973632811
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,1,power_law_1.01,0.20861888885498048
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,1,power_law_1.01,0.2618655967712402
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,1,power_law_1.01,0.31161663055419925
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,1,power_law_1.01,0.42822143554687503
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,1,power_law_1.01,0.5343654251098633
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,1,power_law_1.01,0.7646310424804688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,1,power_law_1.01,0.9785958099365235
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,2048,768,8,128,2,1,power_law_1.01,1.2008025360107422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,1,power_law_1.01,0.036985599994659425
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,1,power_law_1.01,0.043915519714355464
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,1,power_law_1.01,0.0472211217880249
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,1,power_law_1.01,0.052065920829772946
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,1,power_law_1.01,0.039042561054229735
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,1,power_law_1.01,0.06343488216400146
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,1,power_law_1.01,0.06611519813537597
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,1,power_law_1.01,0.06707263946533203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,1,power_law_1.01,0.06844543933868408
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,1,power_law_1.01,0.06837696075439453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,1,power_law_1.01,0.07139391899108886
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,1,power_law_1.01,0.07284800052642823
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,1,power_law_1.01,0.07407423973083496
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,1,power_law_1.01,0.07783232212066651
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,1,power_law_1.01,0.08485695838928223
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,1,power_law_1.01,0.08197376251220703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,1,power_law_1.01,0.09156864166259766
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,1,power_law_1.01,0.10426367759704588
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,1,power_law_1.01,0.12298815727233887
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,1,power_law_1.01,0.15833984375000001
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,1,power_law_1.01,0.19238208770751952
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,1,power_law_1.01,0.2642265510559082
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,1,power_law_1.01,0.3428236770629883
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,1,power_law_1.01,0.5095967864990234
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,1,power_law_1.01,0.6795763397216797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,1,power_law_1.2,0.029699199199676514
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,1,power_law_1.01,1.0176588439941407
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,1,power_law_1.2,0.040640640258789065
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,1,power_law_1.2,0.06536064147949219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,1,power_law_1.2,0.07835711956024169
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,1,power_law_1.01,1.3744473266601562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,1,power_law_1.2,0.12142592430114747
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,1,power_law_1.2,0.12415616035461426
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,1,power_law_1.2,0.12533056259155273
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,1,power_law_1.2,0.1307852840423584
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,1,power_law_1.2,0.13041152000427245
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,1,power_law_1.2,0.13816448211669924
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,1,power_law_1.2,0.1490828800201416
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,1,power_law_1.2,0.15047616004943848
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,1,power_law_1.2,0.15412480354309083
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,1,power_law_1.2,0.15946751594543457
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,1,power_law_1.2,0.15964863777160646
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,1,power_law_1.2,0.17081920623779295
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,1,power_law_1.2,0.20116416931152345
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,1,power_law_1.2,0.2185958480834961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,1,power_law_1.2,0.2688652801513672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,1,power_law_1.2,0.3089254379272461
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,1,power_law_1.2,0.41268737792968746
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,1,power_law_1.2,0.05364096164703369
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,1,power_law_1.2,0.5079935836791992
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2048,768,8,128,2,1,power_law_1.01,1.7331423950195313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,1,power_law_1.2,0.7150816345214844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,1,power_law_1.2,0.8991251373291016
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,1,power_law_1.2,0.03197887897491455
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,1,power_law_1.2,0.0370412802696228
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,1,power_law_1.2,0.046835198402404785
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,1,power_law_1.2,0.053824639320373534
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,1,power_law_1.2,0.07274879932403563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,1,power_law_1.2,0.09752191543579102
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,1,power_law_1.2,0.09913599967956543
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,1,power_law_1.2,0.10163455963134767
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,1,power_law_1.2,0.10281472206115723
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,1,power_law_1.2,0.10292096138000488
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,1,power_law_1.2,0.10416128158569335
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,1,power_law_1.2,0.11835968017578125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,1,power_law_1.2,0.11942720413208008
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,1,power_law_1.2,0.1235142421722412
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,1,power_law_1.2,0.12575488090515136
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,1,power_law_1.2,0.1282444763183594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,1,power_law_1.2,0.13363903999328614
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,1,power_law_1.2,0.14167936325073244
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,1,power_law_1.2,0.15809727668762208
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,1,power_law_1.2,0.1832480049133301
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,1,power_law_1.2,0.20979583740234373
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,1,power_law_1.2,0.2610438346862793
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,1,power_law_1.2,0.31723072052001955
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,1,power_law_1.2,1.2891859436035156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,1,power_law_1.2,0.4265695953369141
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,1,power_law_1.2,0.5362924957275391
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,1,power_law_1.2,0.7590911865234375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,1,power_law_1.2,1.684851837158203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,1,power_law_1.2,0.03734783887863159
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,1,power_law_1.2,0.03989376068115234
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,1,power_law_1.2,0.0438592004776001
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,1,power_law_1.2,0.046441597938537596
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,1,power_law_1.2,0.05086976051330566
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,1,power_law_1.2,0.06429567813873291
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,1,power_law_1.2,0.06637375831604005
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,1,power_law_1.2,0.06799232006072999
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,1,power_law_1.2,0.06935296058654786
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,1,power_law_1.2,0.06991744041442871
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,1,power_law_1.2,0.07197055816650391
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,1,power_law_1.2,0.07448575973510743
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,1,power_law_1.2,0.07622079849243164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,1,power_law_1.2,0.08162624359130859
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,1,power_law_1.2,0.08666560173034668
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,1,power_law_1.2,0.09157504081726074
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2048,768,8,128,2,1,power_law_1.2,2.1226956176757814
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,1,power_law_1.2,0.9799577331542968
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,1,power_law_1.2,0.12535167694091798
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,1,power_law_1.2,0.10108351707458496
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,1,power_law_1.2,0.14932671546936035
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,1,power_law_1.2,0.20515455245971678
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,1,power_law_1.2,0.2607040023803711
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,1,power_law_1.2,0.37958656311035155
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,1,power_law_1.2,0.5049132919311523
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,2048,768,8,128,2,1,power_law_1.2,1.194818572998047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,1,power_law_1.2,0.747158432006836
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,1,power_law_1.2,0.9866783905029297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,1,power_law_1.01,0.0627776002883911
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,1,power_law_1.01,0.08285183906555176
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,1,power_law_1.01,0.12512319564819335
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,1,power_law_1.01,0.1959379196166992
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,1,power_law_1.01,0.23407615661621092
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,1,power_law_1.01,0.3411008071899414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,1,power_law_1.01,0.3621696090698242
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,1,power_law_1.01,0.37029823303222653
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,1,power_law_1.01,0.3925542449951172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,1,power_law_1.01,0.39859905242919924
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,1,power_law_1.01,0.4275462341308594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,1,power_law_1.01,0.44184192657470706
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,1,power_law_1.01,0.444202880859375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,1,power_law_1.01,0.47009727478027347
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,1,power_law_1.01,0.5061676788330078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,1,power_law_1.01,0.5101510238647461
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,1,power_law_1.01,0.5266323089599609
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,1,power_law_1.01,0.5960927963256836
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,1,power_law_1.01,0.6483033752441406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,1,power_law_1.01,0.8480780792236328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,1,power_law_1.01,0.9569631958007813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,1,power_law_1.01,1.3116755676269531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,1,power_law_1.01,1.7950105285644533
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,1,power_law_1.01,2.546611785888672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,1,power_law_1.01,3.2638290405273436
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,1,power_law_1.01,4.762772521972656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,1,power_law_1.2,1.4713381958007812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,1,power_law_1.01,6.374312133789063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,1,power_law_1.01,0.03903808116912842
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,1,power_law_1.01,0.06080319881439209
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,1,power_law_1.01,0.09582528114318847
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,1,power_law_1.01,0.12272128105163574
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,1,power_law_1.01,0.15586048126220703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,1,power_law_1.01,0.2152326393127441
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,1,power_law_1.01,0.23128576278686525
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,2,1,power_law_1.01,8.125305786132813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,1,power_law_1.01,0.23531072616577148
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,1,power_law_1.01,0.24212863922119138
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,1,power_law_1.01,0.2406163215637207
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,1,power_law_1.01,0.24894784927368163
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,1,power_law_1.01,0.2777337646484375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,1,power_law_1.01,0.27848575592041014
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,1,power_law_1.01,0.28476800918579104
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,1,power_law_1.01,0.2921696090698242
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,1,power_law_1.01,0.2975161552429199
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,1,power_law_1.01,0.3120019149780274
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,1,power_law_1.01,0.3297798538208008
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,1,power_law_1.01,0.3550764846801758
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,1,power_law_1.01,0.4124736022949219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,1,power_law_1.01,0.47992702484130856
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,1,power_law_1.01,0.5981843185424804
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,1,power_law_1.01,0.7220575714111328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,1,power_law_1.01,0.9699750518798828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,1,power_law_1.01,1.2409709167480467
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,1,power_law_1.01,1.7832281494140623
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,1,power_law_1.01,2.2957177734375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,2,1,power_law_1.01,2.9087033081054687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,1,power_law_1.01,0.03980544090270996
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,1,power_law_1.01,0.06219456195831299
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,1,power_law_1.01,0.049930877685546875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,1,power_law_1.01,0.07742656230926513
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,1,power_law_1.01,0.0917523193359375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,1,power_law_1.01,0.1273971176147461
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,1,power_law_1.01,0.1359763240814209
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,1,power_law_1.01,0.13866175651550292
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,1,power_law_1.01,0.14417792320251466
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,1,power_law_1.01,0.14437439918518066
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,1,power_law_1.01,0.15176575660705566
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,1,power_law_1.01,0.15457663536071778
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,1,power_law_1.01,0.15876352310180664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,1,power_law_1.01,0.16580991744995116
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,1,power_law_1.2,1.9564620971679687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,1,power_law_1.01,0.18226303100585936
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,1,power_law_1.01,0.17723392486572265
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,1,power_law_1.01,0.19982463836669923
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,1,power_law_1.01,0.2352422332763672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.01,0.27216896057128903
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2048,768,8,128,2,1,power_law_1.2,2.4475199890136716
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.01,0.35167808532714845
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.01,0.43795326232910153
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.01,0.6097683334350587
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.01,0.7920320129394531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.01,1.15279296875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.01,1.51072509765625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,1,power_law_1.2,0.06261760234832764
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,1,power_law_1.2,0.08313599586486817
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,1,power_law_1.2,0.12219008445739746
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,1,power_law_1.2,0.17895872116088868
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,1,power_law_1.2,0.22737024307250975
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,1,power_law_1.2,0.34751296997070313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.01,2.303079071044922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,1,power_law_1.2,0.36442497253417966
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,1,power_law_1.2,0.3894905471801758
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,1,power_law_1.2,0.4014662551879883
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,1,power_law_1.2,0.40863937377929693
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,1,power_law_1.2,0.4332377624511719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,1,power_law_1.2,0.4467417526245117
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,1,power_law_1.2,0.4528441619873047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,1,power_law_1.2,0.4956563186645508
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,1,power_law_1.2,0.5107430267333984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,1,power_law_1.2,0.5179859161376953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,1,power_law_1.2,0.5286399841308593
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,1,power_law_1.2,0.6138924789428711
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,1,power_law_1.2,0.6614534759521484
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,1,power_law_1.2,0.8203327941894532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,1,power_law_1.2,0.9743122863769532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,1,power_law_1.2,1.3654432678222657
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,1,power_law_1.2,1.8255322265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.01,3.08127685546875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,2,1,power_law_1.01,3.8302859497070316
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,1,power_law_1.2,2.5735314941406253
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,1,power_law_1.2,3.2316491699218752
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,1,power_law_1.2,0.04268223762512207
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,1,power_law_1.2,0.06101632118225098
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,1,power_law_1.2,0.0935859203338623
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,1,power_law_1.2,0.11249600410461427
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,1,power_law_1.2,0.14989952087402342
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,1,power_law_1.2,0.21689727783203122
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,1,power_law_1.2,0.23271232604980469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,1,power_law_1.2,0.2386342430114746
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,1,power_law_1.2,0.24410240173339845
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,1,power_law_1.2,0.24331199645996096
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,1,power_law_1.2,0.2475654411315918
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,1,power_law_1.2,0.2786809539794922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,1,power_law_1.2,0.28013311386108397
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,1,power_law_1.2,0.29184383392333985
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,1,power_law_1.2,0.2990713691711426
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,1,power_law_1.2,0.3018227195739746
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,1,power_law_1.2,0.3141907119750976
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,1,power_law_1.2,0.33255168914794925
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,1,power_law_1.2,0.36223678588867186
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,1,power_law_1.2,0.42300991058349613
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,1,power_law_1.2,0.4837478256225586
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,1,power_law_1.2,0.6000012969970703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,1,power_law_1.2,0.723570556640625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,1,power_law_1.2,0.9684697723388671
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,1,power_law_1.2,4.477700500488281
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,1,power_law_1.2,1.225186538696289
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,1,power_law_1.2,5.785164794921875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,1,power_law_1.2,1.739020233154297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,1,power_law_1.2,0.03946176052093506
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,1,power_law_1.2,0.04992063999176025
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,2,1,power_law_1.2,8.066094970703125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,1,power_law_1.2,0.0617145586013794
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,1,power_law_1.2,0.07333951950073242
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,1,power_law_1.2,0.08749567985534668
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,1,power_law_1.2,0.13
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,1,power_law_1.2,0.13711039543151854
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,1,power_law_1.2,0.14258560180664062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,1,power_law_1.2,0.1465932846069336
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,1,power_law_1.2,0.14823552131652834
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,1,power_law_1.2,0.153504638671875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,1,power_law_1.2,0.1601465606689453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,1,power_law_1.2,2.2823980712890624
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,1,power_law_1.2,0.16448575973510743
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,1,power_law_1.2,0.17691007614135743
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,1,power_law_1.2,0.18764095306396483
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,1,power_law_1.2,0.2221811294555664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,1,power_law_1.2,0.27156864166259764
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,1,power_law_1.2,0.1977996826171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.2,0.32847232818603517
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.2,0.447459831237793
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.2,0.5696889495849609
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,2,1,power_law_1.2,2.6591781616210937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.2,0.8315750122070312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.2,1.0952857971191405
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.2,1.6073727416992187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.2,2.1157958984375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,1,power_law_1.01,0.1006668758392334
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,1,power_law_1.01,0.15458239555358885
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,1,power_law_1.01,0.3009753608703613
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,1,power_law_1.01,0.44673088073730466
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,1,power_law_1.01,0.5973376083374023
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,1,power_law_1.01,0.9315968322753907
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,1,power_law_1.01,0.9827072143554687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,1,power_law_1.01,1.020255355834961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,1,power_law_1.01,1.0587622070312501
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,1,power_law_1.01,1.08494140625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,1,power_law_1.01,1.168436508178711
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,1,power_law_1.01,1.256138229370117
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,1,power_law_1.01,1.28329345703125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,1,power_law_1.01,1.3363789367675782
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,1,power_law_1.01,1.3577133178710938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.2,3.15723388671875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,1,power_law_1.01,1.3777197265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,1,power_law_1.01,1.4982681274414062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,1,power_law_1.01,1.7806906127929687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,1,power_law_1.01,2.081301727294922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,1,power_law_1.01,2.441031646728516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,1,power_law_1.01,3.093201293945312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,1,power_law_1.01,3.7989471435546873
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,1,power_law_1.01,4.911947631835938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,1,power_law_1.01,6.410535888671875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,1,power_law_1.01,9.1980322265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,1,power_law_1.01,12.319847412109375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.2,4.194603576660156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,1,power_law_1.01,0.07161727905273438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,1,power_law_1.01,0.11281023979187013
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,1,power_law_1.01,0.17387136459350586
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,1,power_law_1.01,0.24930303573608398
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,1,power_law_1.01,0.334021110534668
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,1,power_law_1.01,0.5104844665527344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,1,power_law_1.01,0.5399411010742188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,1,power_law_1.01,16.748273925781252
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,1,power_law_1.01,0.5583552169799805
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,1,power_law_1.01,0.5788025665283203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,1,power_law_1.01,0.5817222213745117
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,1,power_law_1.01,0.6004121780395508
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,1,power_law_1.01,0.6152441787719727
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,1,power_law_1.01,0.6873241424560547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,1,power_law_1.01,0.7032243347167968
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,1,power_law_1.01,0.7189740753173828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,1,power_law_1.01,0.6848767852783203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,1,power_law_1.01,0.7464031982421875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,1,power_law_1.01,0.8039981079101561
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,1,power_law_1.01,0.8252236938476563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,1,power_law_1.01,0.9579571533203126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,1,power_law_1.01,1.224666213989258
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,2,1,power_law_1.01,19.52077880859375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,1,power_law_1.01,1.4831289672851562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,1,power_law_1.01,1.8121253967285156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,1,power_law_1.01,2.3511616516113283
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,1,power_law_1.01,3.057661437988281
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,1,power_law_1.01,4.3131845092773435
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,1,power_law_1.01,5.170833129882813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,2,1,power_law_1.01,6.794085083007812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,1,power_law_1.01,0.06369408130645751
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,1,power_law_1.01,0.15049023628234864
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,1,power_law_1.01,0.1952672004699707
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,1,power_law_1.01,0.1071571159362793
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,1,power_law_1.01,0.3041292762756348
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,1,power_law_1.01,0.08027775764465332
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,1,power_law_1.01,0.3199443244934082
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,1,power_law_1.01,0.3339807891845703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,1,power_law_1.01,0.3446547317504883
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,1,power_law_1.01,0.3491167831420898
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,1,power_law_1.01,0.362229118347168
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,1,power_law_1.01,0.37207359313964844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,1,power_law_1.01,0.37615745544433593
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,1,power_law_1.01,0.3986739349365235
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,1,power_law_1.01,0.4080217742919922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,1,power_law_1.01,0.41165695190429685
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,2,1,power_law_1.2,5.195779418945312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.01,0.5610559844970704
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,1,power_law_1.01,0.4465107345581055
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,1,power_law_1.01,0.5083852767944336
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.01,0.6857778930664062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.01,0.8406848144531249
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.01,1.1153651428222655
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.01,1.3848159790039063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.01,1.989141082763672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.01,2.5840231323242184
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,1,power_law_1.2,0.10130111694335937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,1,power_law_1.2,0.15613439559936523
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,1,power_law_1.2,0.29298368453979495
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.01,3.8387322998046876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,1,power_law_1.2,0.40072063446044925
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,1,power_law_1.2,0.9188256072998048
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,1,power_law_1.2,0.5478252792358399
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,1,power_law_1.2,0.9690879821777344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,1,power_law_1.2,1.0330303955078126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,1,power_law_1.2,1.0856511688232422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,1,power_law_1.2,1.144610595703125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,1,power_law_1.2,1.243281936645508
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,1,power_law_1.2,1.3946330261230468
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,1,power_law_1.2,1.329471435546875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,1,power_law_1.2,1.351041259765625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,1,power_law_1.2,1.4284410095214843
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,1,power_law_1.2,1.4645344543457033
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,1,power_law_1.2,1.5450912475585938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.01,5.0971789550781255
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,1,power_law_1.2,1.8467187500000002
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,1,power_law_1.2,2.187279357910156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,1,power_law_1.2,2.4846456909179686
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,1,power_law_1.2,3.0959487915039063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,1,power_law_1.2,4.125014343261719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,1,power_law_1.2,4.626653442382812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,1,power_law_1.2,6.955575561523437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,1,power_law_1.2,8.940798339843749
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,2560,8,160,2,1,power_law_1.01,6.393336181640625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,1,power_law_1.2,0.0728991985321045
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,1,power_law_1.2,0.11383551597595214
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,1,power_law_1.2,0.16746944427490235
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,1,power_law_1.2,0.2345907211303711
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,1,power_law_1.2,0.3211065673828125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,1,power_law_1.2,0.5160063934326172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,1,power_law_1.2,12.130025634765625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,1,power_law_1.2,0.5356256103515624
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,1,power_law_1.2,0.5632652664184571
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,1,power_law_1.2,0.5844179153442383
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,1,power_law_1.2,0.5963647842407227
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,1,power_law_1.2,0.6117798233032227
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,1,power_law_1.2,0.6193094253540039
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,1,power_law_1.2,0.694750747680664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,1,power_law_1.2,0.6999104309082032
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,1,power_law_1.2,0.7207353973388673
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,1,power_law_1.2,0.7286144256591797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,1,power_law_1.2,0.7545676422119141
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,1,power_law_1.2,0.8073260498046875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,1,power_law_1.2,0.8590220642089843
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,1,power_law_1.2,0.9811917114257813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,1,power_law_1.2,1.1461580657958985
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,1,power_law_1.2,1.4301759338378905
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,1,power_law_1.2,1.7420953369140626
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,1,power_law_1.2,2.29801025390625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,1,power_law_1.2,2.867850341796875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,1,power_law_1.2,16.61767578125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,1,power_law_1.2,4.060584411621094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,1,power_law_1.2,0.06453375816345215
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,1,power_law_1.2,0.07634751796722412
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,1,power_law_1.2,0.10382975578308104
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,1,power_law_1.2,0.13781696319580078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,1,power_law_1.2,0.18150144577026367
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,1,power_law_1.2,0.2986483192443848
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,1,power_law_1.2,0.31744319915771485
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,1,power_law_1.2,0.33896766662597655
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,1,power_law_1.2,0.35014465332031247
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,2,1,power_law_1.2,20.075689697265624
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,1,power_law_1.2,0.35963199615478514
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,1,power_law_1.2,0.37371776580810545
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,1,power_law_1.2,4.7612677001953125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,1,power_law_1.2,0.3801824188232422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,1,power_law_1.2,0.39546558380126956
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,2,1,power_law_1.2,6.62219970703125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,1,power_law_1.2,0.4511667251586914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,1,power_law_1.2,0.43195648193359376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,1,power_law_1.2,0.40927616119384763
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,1,power_law_1.2,0.5712275314331055
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,1,power_law_1.2,0.49251712799072267
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.2,0.8539622497558593
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.2,1.0615321350097655
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.2,0.6551148986816406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.2,1.52015869140625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.2,1.9848793029785157
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.2,2.9716589355468748
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.2,3.98724853515625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,1,power_law_1.01,0.10346112251281739
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,1,power_law_1.01,0.14451135635375975
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,1,power_law_1.01,0.28371904373168944
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,1,power_law_1.01,0.4912607955932617
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,1,power_law_1.01,0.8477555084228516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,1,power_law_1.01,1.1670713806152344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,1,power_law_1.01,1.3528530883789063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,1,power_law_1.01,1.753395233154297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,1,power_law_1.01,1.8484640502929686
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,1,power_law_1.01,1.949585876464844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,1,power_law_1.01,2.026291809082031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,1,power_law_1.01,2.060985565185547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,1,power_law_1.01,2.2154981994628904
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,1,power_law_1.01,2.409743347167969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.2,6.068345947265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,1,power_law_1.01,2.5913882446289063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,1,power_law_1.01,2.919530334472656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,1,power_law_1.01,2.5814968872070314
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,1,power_law_1.01,3.0013189697265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,1,power_law_1.01,3.62844482421875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,1,power_law_1.01,4.32344970703125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,1,power_law_1.01,4.962279663085938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,1,power_law_1.01,6.398519287109375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,1,power_law_1.01,8.057561645507812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.2,8.090755004882812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,1,power_law_1.01,10.590677490234375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,1,power_law_1.01,13.328605957031249
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,1,power_law_1.01,0.07158143997192383
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,1,power_law_1.01,0.09671168327331543
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,1,power_law_1.01,0.15958784103393556
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,1,power_law_1.01,0.28009471893310545
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,1,power_law_1.01,0.47097534179687506
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,1,power_law_1.01,16.428878173828124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,1,power_law_1.01,0.6579551696777344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,1,power_law_1.01,0.7826758575439453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,1,power_law_1.01,0.967992935180664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,1,power_law_1.01,1.020016632080078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,1,power_law_1.01,1.0568787384033205
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,1,power_law_1.01,1.093117446899414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,1,power_law_1.01,1.1083769226074218
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,2,1,power_law_1.01,21.42982177734375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,1,power_law_1.01,1.1400077056884765
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,1,power_law_1.01,1.1913785552978515
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,1,power_law_1.01,1.2040589141845703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,1,power_law_1.01,1.2438861083984374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,1,power_law_1.01,1.3987992858886718
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,1,power_law_1.01,1.4571405029296876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,1,power_law_1.01,1.5264019775390625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,1,power_law_1.01,1.613477783203125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,1,power_law_1.01,2.0715263366699217
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,1,power_law_1.01,2.086901092529297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,1,power_law_1.01,2.432508087158203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,1,power_law_1.01,3.253895568847656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,1,power_law_1.01,3.5557876586914063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,1,power_law_1.01,4.782126770019532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,1,power_law_1.01,6.62138671875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,2,1,power_law_1.01,7.008424072265624
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,1,power_law_1.01,0.08681728363037108
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,1,power_law_1.01,0.0878758430480957
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,1,power_law_1.01,0.09849087715148926
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,1,power_law_1.01,0.11160384178161623
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,1,power_law_1.01,0.1238265609741211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,1,power_law_1.01,0.137324800491333
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,1,power_law_1.01,0.08936832427978517
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,1,power_law_1.01,0.14934144020080567
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,1,power_law_1.01,0.16245695114135744
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,1,power_law_1.01,0.18620031356811523
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,1,power_law_1.01,0.08931327819824218
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,1,power_law_1.01,0.22331264495849606
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,1,power_law_1.01,0.24959039688110352
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,1,power_law_1.01,0.2980179214477539
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,1,power_law_1.01,0.3633203125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,1,power_law_1.01,0.41257472991943356
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,1,power_law_1.01,0.536924171447754
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,1,power_law_1.01,0.7842310333251954
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,1,power_law_1.01,1.0197427368164063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,1,power_law_1.01,1.4910713195800782
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,1,power_law_1.01,1.9618885803222657
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,1,power_law_1.01,2.891893615722656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,1,power_law_1.01,3.830072937011719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,1,power_law_1.01,5.701375732421875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,1,power_law_1.2,0.15552703857421873
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,1,power_law_1.2,0.28427072525024416
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,1,power_law_1.2,0.49066303253173826
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,1,power_law_1.2,0.7618303680419922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,1,power_law_1.2,1.0705606079101562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,1,power_law_1.2,1.2429535675048828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,1,power_law_1.2,1.704094696044922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,1,power_law_1.2,1.806953582763672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,1,power_law_1.2,1.8541952514648437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,1,power_law_1.2,1.9236805725097657
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,1,power_law_1.2,2.0625433349609374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,1,power_law_1.2,2.202112579345703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,1,power_law_1.2,2.4276780700683593
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,1,power_law_1.2,2.527892456054688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,1,power_law_1.2,2.735631408691406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,1,power_law_1.2,2.92418701171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,1,power_law_1.2,2.704364013671875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,1,power_law_1.2,3.269962768554687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,1,power_law_1.2,3.939980773925781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,1,power_law_1.2,4.380215759277344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,1,power_law_1.2,5.44481689453125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,1,power_law_1.2,6.1853399658203125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,1,power_law_1.2,8.712638549804687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,1,power_law_1.2,10.676585693359375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,1,power_law_1.2,13.264707031250001
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,1,power_law_1.2,16.660931396484376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,1,power_law_1.2,0.0672979211807251
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,1,power_law_1.2,0.15873663902282714
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,1,power_law_1.2,0.2797990417480469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,1,power_law_1.2,0.4284812927246094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,1,power_law_1.2,0.09284799575805665
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,1,power_law_1.2,0.6249081420898437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,1,power_law_1.2,0.7252313232421874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,1,power_law_1.2,0.9526137542724609
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,1,power_law_1.2,1.0083404541015626
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,1,power_law_1.2,1.0237510681152344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,1,power_law_1.2,1.055438690185547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,1,power_law_1.2,1.1094700622558595
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,1,power_law_1.2,1.1348767852783204
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,1,power_law_1.2,1.1717772674560547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,1,power_law_1.2,1.2225823974609376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,1,power_law_1.2,1.2189625549316407
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,2,1,power_law_1.2,21.52037841796875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,1,power_law_1.2,1.3800128173828123
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,1,power_law_1.2,1.5073081970214843
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,1,power_law_1.2,1.6248512268066406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,1,power_law_1.2,1.9133094787597655
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,1,power_law_1.2,1.8912390136718749
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,1,power_law_1.2,2.351026611328125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,1,power_law_1.2,2.716882019042969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,1,power_law_1.2,3.1845733642578127
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,1,power_law_1.2,4.113897094726562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,1,power_law_1.2,4.738737792968751
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,1,power_law_1.2,6.30165283203125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,1,power_law_1.2,0.08743552207946778
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,1,power_law_1.2,0.09203200340270996
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,2,1,power_law_1.2,7.847880249023437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,1,power_law_1.2,0.09855168342590333
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,1,power_law_1.2,0.12369536399841308
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,1,power_law_1.2,0.13633536338806151
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,1,power_law_1.2,0.14907584190368653
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,1,power_law_1.2,0.16136640548706055
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,1,power_law_1.2,0.18600576400756835
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,1,power_law_1.2,0.22396095275878905
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,1,power_law_1.2,0.2487763214111328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,1,power_law_1.2,0.08641535758972167
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,1,power_law_1.2,0.2996326446533203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,1,power_law_1.2,0.4137926483154297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,1,power_law_1.2,0.08847231864929199
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,1,power_law_1.2,0.1113657569885254
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,1,power_law_1.2,0.3612953567504883
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,1,power_law_1.2,0.7857087707519531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,1,power_law_1.2,0.5441843032836914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,1,power_law_1.2,1.0223782348632813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,1,power_law_1.2,1.4879238891601563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,1,power_law_1.2,1.9624671936035156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,1,power_law_1.2,2.8912921142578125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,1,power_law_1.2,3.8439205932617186
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,1,power_law_1.2,5.698177490234375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,1,power_law_1.01,0.07657855987548828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,1,power_law_1.01,0.08896384239196778
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,1,power_law_1.01,0.14704319953918457
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,1,power_law_1.01,0.17564416885375977
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,1,power_law_1.01,0.26276479721069335
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,1,power_law_1.01,0.2792947196960449
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,1,power_law_1.01,0.29383871078491214
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,1,power_law_1.01,0.3020166397094727
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,1,power_law_1.01,0.3070444869995117
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,1,power_law_1.01,0.32367744445800783
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,1,power_law_1.01,0.36040191650390624
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,1,power_law_1.01,0.37198848724365235
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,1,power_law_1.01,0.3799308776855469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,1,power_law_1.01,0.3797747039794922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,1,power_law_1.01,0.3834860610961914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,1,power_law_1.01,0.4015135955810547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,1,power_law_1.01,0.4453388977050781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,1,power_law_1.01,0.479541130065918
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,1,power_law_1.01,0.5610892868041992
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,1,power_law_1.01,0.6541375732421875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,1,power_law_1.01,0.8422841644287109
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,1,power_law_1.01,1.0257350158691405
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,1,power_law_1.01,1.5103974914550782
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,1,power_law_1.01,1.9164883422851564
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,1,power_law_1.01,2.7598699951171874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,2,1,power_law_1.01,4.501537170410156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,1,power_law_1.01,3.7431085205078127
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,1,power_law_1.01,0.039064319133758546
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,1,power_law_1.01,0.08129983901977539
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,1,power_law_1.01,0.04359551906585694
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,1,power_law_1.01,0.10795328140258789
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,1,power_law_1.01,0.10980607986450194
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,1,power_law_1.01,0.11758720397949218
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,1,power_law_1.01,0.12052288055419921
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,1,power_law_1.01,0.12214847564697266
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,1,power_law_1.01,0.1250361633300781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,1,power_law_1.01,0.1290227222442627
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,1,power_law_1.01,0.1323526382446289
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,1,power_law_1.01,0.13681920051574709
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,1,power_law_1.01,0.14073408126831055
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,1,power_law_1.01,0.14549504280090333
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,1,power_law_1.01,0.05330815792083741
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,1,power_law_1.01,0.06707263946533203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,1,power_law_1.01,0.1795577621459961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,1,power_law_1.01,0.1571782398223877
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,1,power_law_1.01,0.2008243179321289
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,1,power_law_1.01,0.2544416046142578
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,1,power_law_1.01,0.3073376083374023
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,1,power_law_1.01,0.4150796890258789
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,1,power_law_1.01,0.5244512176513672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,1,power_law_1.01,0.7567833709716797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,1,power_law_1.01,1.001007385253906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,1,power_law_1.01,1.4655052185058595
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,1,power_law_1.2,0.044199042320251465
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,1,power_law_1.2,0.08709247589111328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,1,power_law_1.01,1.9533293151855466
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,1,power_law_1.2,0.13058367729187012
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,1,power_law_1.2,0.2613196754455566
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,1,power_law_1.2,0.28181440353393555
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,1,power_law_1.2,0.2961356735229492
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,1,power_law_1.2,0.05884223937988281
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,1,power_law_1.2,0.3030496025085449
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2688,1856,6,128,2,1,power_law_1.01,2.4840109252929685
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,1,power_law_1.2,0.3097043228149414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,1,power_law_1.2,0.331644172668457
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,1,power_law_1.2,0.3771750259399414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,1,power_law_1.2,0.38476478576660156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,1,power_law_1.2,0.3841548919677734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,1,power_law_1.2,0.37404033660888675
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,1,power_law_1.2,0.39085887908935546
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,1,power_law_1.2,0.40987071990966795
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,1,power_law_1.2,0.4501126480102539
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,1,power_law_1.2,0.4859667205810547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,1,power_law_1.2,0.1669887924194336
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,1,power_law_1.2,0.5803532791137695
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,1,power_law_1.2,0.6753683471679688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,1,power_law_1.2,0.8562028503417969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,1,power_law_1.2,1.0556050872802734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,1,power_law_1.2,1.5271827697753906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,1,power_law_1.2,2.017587127685547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,1,power_law_1.2,0.0390233588218689
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,1,power_law_1.2,0.045966720581054686
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,1,power_law_1.2,0.05489727973937988
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,1,power_law_1.2,0.06475456237792969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,1,power_law_1.2,2.762261657714844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,1,power_law_1.2,0.07755136013031005
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,1,power_law_1.2,0.10773823738098146
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,1,power_law_1.2,0.11330176353454591
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,1,power_law_1.2,0.11876416206359863
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,1,power_law_1.2,0.1212435245513916
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,1,power_law_1.2,0.12383808135986327
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,1,power_law_1.2,0.12746047973632812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,1,power_law_1.2,0.13319680213928223
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,1,power_law_1.2,0.13662336349487306
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,1,power_law_1.2,0.14339648246765138
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,1,power_law_1.2,0.15089728355407714
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,1,power_law_1.2,3.5422994995117185
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,1,power_law_1.2,0.2041548728942871
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,2,1,power_law_1.2,4.695419616699219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,1,power_law_1.2,0.15728832244873048
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,1,power_law_1.2,0.1712825584411621
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,1,power_law_1.2,0.23779903411865236
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,1,power_law_1.2,0.3200204849243164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,1,power_law_1.2,0.3991551971435547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,1,power_law_1.2,0.5733843231201172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,1,power_law_1.2,0.7534547424316406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,1,power_law_1.2,1.1325516510009765
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,1,power_law_1.2,1.5219731140136719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,1,power_law_1.01,0.17909568786621094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,1,power_law_1.01,0.28035903930664063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,1,power_law_1.2,2.340975341796875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,1,power_law_1.01,0.4986342239379883
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,1,power_law_1.01,0.834234848022461
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,1,power_law_1.01,1.1632543945312501
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,1,power_law_1.01,1.9644371032714845
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,1,power_law_1.01,2.0730393981933597
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,1,power_law_1.01,2.3339334106445313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,1,power_law_1.01,2.2428620910644534
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,1,power_law_1.01,2.446787872314453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,1,power_law_1.2,3.148901672363281
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,1,power_law_1.01,2.5631860351562503
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,1,power_law_1.01,2.931548767089844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,1,power_law_1.01,2.889281921386719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,1,power_law_1.01,3.1540646362304687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,1,power_law_1.01,3.8304229736328126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,1,power_law_1.01,3.8196844482421874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,1,power_law_1.01,4.384600219726563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,1,power_law_1.01,4.124543457031249
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,1,power_law_1.01,4.287434387207031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,1,power_law_1.01,5.165143432617187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,1,power_law_1.01,6.849485473632813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2688,1856,6,128,2,1,power_law_1.2,3.946707763671875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,1,power_law_1.01,8.388347778320313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,1,power_law_1.01,10.3183251953125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,1,power_law_1.01,14.566663818359377
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,1,power_law_1.01,18.93764404296875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,1,power_law_1.01,24.19644287109375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,1,power_law_1.01,0.11311231613159181
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,1,power_law_1.01,0.17751808166503907
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,1,power_law_1.01,0.29662656784057617
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,1,power_law_1.01,30.624392089843752
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,1,power_law_1.01,0.4792480087280273
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,1,power_law_1.01,0.6844364929199218
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,1,power_law_1.01,1.0791871643066406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,1,power_law_1.01,1.1480902099609376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,1,power_law_1.01,1.2484518432617189
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,1,power_law_1.01,1.2581228637695312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,1,power_law_1.01,1.2360192108154298
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,1,power_law_1.01,1.294082489013672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,1,power_law_1.01,1.3301248168945312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,1,power_law_1.01,1.3810861206054688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,1,power_law_1.01,1.4112416076660157
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,1,power_law_1.01,1.4242323303222657
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,1,power_law_1.01,1.5634393310546875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,1,power_law_1.01,1.5762911987304686
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,2,1,power_law_1.01,40.34138427734375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,1,power_law_1.01,1.8539213562011718
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,1,power_law_1.01,2.1755039978027346
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,1,power_law_1.01,2.342913970947266
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,1,power_law_1.01,3.3368453979492188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,1,power_law_1.01,2.878540649414062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,1,power_law_1.01,4.3568353271484375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,1,power_law_1.01,5.142677612304688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,1,power_law_1.01,6.540119018554688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,1,power_law_1.01,9.885679321289063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,1,power_law_1.01,12.851967773437499
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,2,1,power_law_1.01,14.1998486328125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,1,power_law_1.01,0.12905280113220213
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,1,power_law_1.01,0.1303135967254639
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,1,power_law_1.01,0.13823360443115235
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,1,power_law_1.01,0.15160191535949707
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,1,power_law_1.01,0.16205440521240236
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,1,power_law_1.01,0.13043519973754883
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,1,power_law_1.01,0.1729132843017578
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,1,power_law_1.01,0.18488447189331053
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,1,power_law_1.01,0.1969049644470215
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,1,power_law_1.01,0.21970495223999023
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,1,power_law_1.01,0.1313152027130127
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,1,power_law_1.01,0.2654278373718262
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,1,power_law_1.01,0.3363692855834961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,1,power_law_1.01,0.2916281509399414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,1,power_law_1.01,0.4057785415649414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,1,power_law_1.01,0.4502540969848633
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,1,power_law_1.01,0.5709996795654296
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,1,power_law_1.01,1.0402738952636719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,1,power_law_1.01,0.8045772552490235
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,1,power_law_1.01,1.5045529174804688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,1,power_law_1.01,1.9776448059082032
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,1,power_law_1.01,2.902349548339844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,1,power_law_1.01,3.8478329467773436
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,1,power_law_1.01,5.676488037109375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,1,power_law_1.2,0.1810028839111328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,1,power_law_1.01,7.53558837890625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,1,power_law_1.2,0.2811084747314453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,1,power_law_1.2,0.7654918670654297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,1,power_law_1.2,0.4964684677124024
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,1,power_law_1.2,1.0376172637939454
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,1,power_law_1.2,1.9772741699218748
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,1,power_law_1.2,2.061670989990234
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,1,power_law_1.2,2.314019775390625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,1,power_law_1.2,2.3298272705078125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,1,power_law_1.2,2.6120947265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,1,power_law_1.2,2.630504455566406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,1,power_law_1.2,3.0666009521484376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,1,power_law_1.2,3.031923828125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,1,power_law_1.2,3.574429931640625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,1,power_law_1.2,3.65196533203125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,1,power_law_1.2,4.2075936889648435
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,1,power_law_1.2,4.493696594238282
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,1,power_law_1.2,4.326394958496094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,1,power_law_1.2,4.562304077148438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,1,power_law_1.2,5.711714477539063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,1,power_law_1.2,6.3172888183593745
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,1,power_law_1.2,9.187830810546874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,1,power_law_1.2,10.25447998046875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,1,power_law_1.2,14.651895751953125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,1,power_law_1.2,16.98069091796875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,1,power_law_1.2,0.11110848426818846
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,1,power_law_1.2,0.17420480728149415
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,1,power_law_1.2,0.2933235168457031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,1,power_law_1.2,0.4405427169799805
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,1,power_law_1.2,0.6127718353271485
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,1,power_law_1.2,1.0621683502197266
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,1,power_law_1.2,1.1511641693115233
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,1,power_law_1.2,1.19106689453125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,1,power_law_1.2,1.2442540740966797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,1,power_law_1.2,1.276904296875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,1,power_law_1.2,1.316887664794922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,1,power_law_1.2,1.363186492919922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,1,power_law_1.2,1.3810719299316407
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,1,power_law_1.2,1.4675456237792968
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,1,power_law_1.2,1.50403076171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,1,power_law_1.2,25.813137207031254
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,1,power_law_1.2,1.5600679016113281
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,1,power_law_1.2,2.020016632080078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,1,power_law_1.2,1.7152793884277344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,1,power_law_1.2,2.1270501708984377
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,1,power_law_1.2,2.5443353271484375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,1,power_law_1.2,2.7583041381835938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,1,power_law_1.2,3.5776107788085936
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,1,power_law_1.2,4.008150939941406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,1,power_law_1.2,5.377734985351562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,1,power_law_1.2,6.397880859375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,1,power_law_1.2,30.592402343750003
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,1,power_law_1.2,9.67013427734375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,1,power_law_1.2,0.12902527809143066
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,1,power_law_1.2,0.1289299201965332
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,1,power_law_1.2,0.12921279907226563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,1,power_law_1.2,0.13259263992309572
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,1,power_law_1.2,0.1379430389404297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,1,power_law_1.2,0.1512838363647461
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,1,power_law_1.2,0.1619308853149414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,2,1,power_law_1.2,40.106689453125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,1,power_law_1.2,0.17383935928344726
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,1,power_law_1.2,0.19761856079101564
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,1,power_law_1.2,11.699061279296874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,1,power_law_1.2,0.21984064102172854
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,1,power_law_1.2,0.2660326385498047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,1,power_law_1.2,0.2898828887939453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,1,power_law_1.2,0.3361017608642578
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,1,power_law_1.2,0.45187198638916015
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,1,power_law_1.2,0.18490688323974608
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,1,power_law_1.2,0.40592384338378906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,2,1,power_law_1.2,14.009916992187499
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,1,power_law_1.2,0.5686355209350585
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,1,power_law_1.2,0.8041580963134767
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,1,power_law_1.2,1.0414905548095703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,1,power_law_1.2,1.5028262329101563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,1,power_law_1.2,1.9741754150390627
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,1,power_law_1.2,2.9160894775390624
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,1,power_law_1.2,3.8366424560546877
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,1,power_law_1.2,5.676823120117187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,1,power_law_1.01,0.08978495597839356
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,1,power_law_1.01,0.0965062427520752
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,1,power_law_1.01,0.10801088333129882
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,1,power_law_1.01,0.14810367584228518
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,1,power_law_1.01,0.17480127334594725
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,1,power_law_1.01,0.17580032348632812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,1,power_law_1.01,0.17719871520996094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,1,power_law_1.01,0.17645887374877928
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,1,power_law_1.01,0.17684032440185546
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,1,power_law_1.01,0.18619840621948242
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,1,power_law_1.01,0.22303104400634766
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,1,power_law_1.01,0.22879871368408203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,1,power_law_1.01,0.23757375717163084
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,1,power_law_1.01,0.25088575363159177
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,1,power_law_1.01,0.2592972755432129
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,1,power_law_1.01,0.30038976669311523
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,1,power_law_1.01,0.3485670471191406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,1,power_law_1.01,0.42276031494140626
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,1,power_law_1.01,0.5043564987182617
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,1,power_law_1.2,7.5189111328125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,1,power_law_1.01,0.6762873840332031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,1,power_law_1.01,0.830346908569336
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,1,power_law_1.01,1.2381337738037108
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,1,power_law_1.01,1.597966766357422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,1,power_law_1.01,2.4332492065429685
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,1,power_law_1.01,3.1302163696289065
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,1,power_law_1.01,4.8056665039062505
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,1,power_law_1.01,0.04809535980224609
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,1,power_law_1.01,0.059822077751159664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,1,power_law_1.01,0.06504703998565674
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,1,power_law_1.01,0.09194111824035645
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,1,power_law_1.01,0.09691647529602052
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,1,power_law_1.01,0.09870719909667969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,1,power_law_1.01,0.10151231765747071
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,1,power_law_1.01,0.10247872352600099
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,1,power_law_1.01,0.10171008110046387
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,1,power_law_1.01,0.10428735733032227
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,1,power_law_1.01,0.10496064186096192
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,1,power_law_1.01,0.10563008308410644
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,1,power_law_1.01,0.10793343544006348
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,1,power_law_1.01,0.11971263885498047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,1,power_law_1.01,6.363326416015625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,1,power_law_1.01,0.12726783752441406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,1,power_law_1.01,0.13977279663085937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,1,power_law_1.01,0.16265151977539063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,1,power_law_1.01,0.19166080474853514
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,1,power_law_1.01,0.13598527908325195
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,1,power_law_1.01,0.24628671646118164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,1,power_law_1.01,0.2818086433410644
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,1,power_law_1.01,0.3930329513549805
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,1,power_law_1.01,0.4937958526611328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,1,power_law_1.01,0.7167839813232422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,1,power_law_1.01,0.924956817626953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,4,1,power_law_1.01,8.412301635742187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,1,power_law_1.01,1.4482681274414062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,1,power_law_1.01,1.9924972534179688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,4,1,power_law_1.01,0.026331520080566405
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,4,1,power_law_1.01,0.05126016139984131
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,4,1,power_law_1.01,0.05162752151489258
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,4,1,power_law_1.01,0.052133121490478516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,4,1,power_law_1.01,0.0521619176864624
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,4,1,power_law_1.01,0.05200511932373046
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,4,1,power_law_1.01,0.051675519943237304
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,4,1,power_law_1.01,0.05227583885192871
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,4,1,power_law_1.01,0.052562561035156254
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,4,1,power_law_1.01,2.37901123046875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,4,1,power_law_1.01,0.034904320240020756
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,4,1,power_law_1.01,0.038578560352325444
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,4,1,power_law_1.01,0.05387328147888184
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,4,1,power_law_1.01,0.06111167907714844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,4,1,power_law_1.01,0.05572671890258789
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,4,1,power_law_1.01,0.06089024066925049
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,4,1,power_law_1.01,0.06675456047058106
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,4,1,power_law_1.01,0.07305408000946045
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,4,1,power_law_1.01,0.08932991981506347
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.01,0.1058944034576416
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.01,0.13783743858337402
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.01,0.1733292770385742
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.01,0.23960512161254882
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.01,0.3079084777832031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.01,0.447534065246582
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.01,0.5891654586791992
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,1,power_law_1.2,0.08036479949951172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,1,power_law_1.2,0.09326335906982422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,1,power_law_1.2,0.10035264015197753
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,1,power_law_1.2,0.1477939224243164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,1,power_law_1.2,0.1713145637512207
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,1,power_law_1.2,0.17607103347778322
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,1,power_law_1.2,0.17750335693359376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,1,power_law_1.2,0.17908544540405275
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,1,power_law_1.2,0.18106176376342772
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,1,power_law_1.2,0.18764736175537108
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,1,power_law_1.2,0.22729087829589845
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,1,power_law_1.2,0.23153856277465817
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,1,power_law_1.2,0.2362700843811035
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,1,power_law_1.2,0.2504870414733887
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,1,power_law_1.2,0.26035200119018553
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,1,power_law_1.2,0.3117318344116211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,1,power_law_1.2,0.3485932922363281
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,1,power_law_1.2,0.4241139221191406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,1,power_law_1.2,0.5115622329711914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,1,power_law_1.2,0.6825523376464844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,1,power_law_1.2,0.846221466064453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,1,power_law_1.2,1.2195929718017577
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.01,0.8720108795166016
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.01,1.1421075439453126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,1,power_law_1.2,1.6605267333984375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,1,power_law_1.2,2.4083544921874998
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,4,1,power_law_1.01,1.4327078247070313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,1,power_law_1.2,3.360470275878906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,1,power_law_1.2,0.05268288135528565
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,1,power_law_1.2,0.06282944202423095
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,1,power_law_1.2,0.06523839950561523
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,1,power_law_1.2,0.09156352043151855
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,1,power_law_1.2,0.0953651237487793
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,1,power_law_1.2,0.09734592437744141
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,1,power_law_1.2,0.0984659194946289
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,1,power_law_1.2,0.10011648178100585
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,1,power_law_1.2,0.10046079635620117
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,1,power_law_1.2,0.10258624076843262
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,1,power_law_1.2,0.10414527893066405
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,1,power_law_1.2,0.10602751731872559
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,1,power_law_1.2,0.10576704025268555
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,1,power_law_1.2,0.11763456344604492
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,1,power_law_1.2,0.12942272186279297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,1,power_law_1.2,0.14145919799804688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,1,power_law_1.2,0.13805055618286133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,1,power_law_1.2,0.1586246395111084
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,1,power_law_1.2,0.18978559494018554
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,1,power_law_1.2,0.23403263092041016
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,1,power_law_1.2,0.2826764869689941
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,1,power_law_1.2,5.018044128417968
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,1,power_law_1.2,0.3914739227294922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,1,power_law_1.2,0.49926208496093755
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,1,power_law_1.2,0.721045150756836
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,1,power_law_1.2,0.9592697906494141
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,1,power_law_1.2,6.370992431640625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,4,1,power_law_1.2,8.386631469726563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,4,1,power_law_1.2,0.026009600162506103
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,4,1,power_law_1.2,0.03412287950515747
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,4,1,power_law_1.2,0.0364083194732666
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,4,1,power_law_1.2,0.05114687919616699
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,4,1,power_law_1.2,0.050588798522949216
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,4,1,power_law_1.2,0.052195200920104975
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,4,1,power_law_1.2,0.0522540807723999
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,4,1,power_law_1.2,0.05170623779296875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,4,1,power_law_1.2,0.05228288173675537
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,4,1,power_law_1.2,0.052133121490478516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,4,1,power_law_1.2,0.052988162040710454
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,4,1,power_law_1.2,0.05407872200012207
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,4,1,power_law_1.2,0.055831041336059574
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,4,1,power_law_1.2,0.060339841842651364
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,4,1,power_law_1.2,0.06437952041625976
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,1,power_law_1.2,1.4938265991210937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,4,1,power_law_1.2,0.06894976139068604
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,4,1,power_law_1.2,0.07270592212677002
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,4,1,power_law_1.2,0.09112832069396973
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.2,0.10757504463195802
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.2,0.14002559661865235
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,1,power_law_1.2,1.8698335266113282
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.2,0.1758995246887207
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.2,0.2505881690979004
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,4,1,power_law_1.2,2.3842477416992187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.2,0.32133438110351564
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.2,0.46301376342773437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.2,0.8899987030029297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,1,power_law_1.01,0.1028371238708496
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,1,power_law_1.01,0.13006912231445314
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,1,power_law_1.01,0.15604991912841798
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,1,power_law_1.01,0.23958911895751953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,1,power_law_1.01,0.24938751220703126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,1,power_law_1.01,0.2508473587036133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,1,power_law_1.01,0.25165311813354496
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,1,power_law_1.01,0.2542259216308594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,1,power_law_1.01,0.25275392532348634
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,1,power_law_1.01,0.27298559188842775
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,1,power_law_1.01,0.34296321868896484
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.2,0.6067667388916016
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,1,power_law_1.01,0.3640166473388672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,1,power_law_1.01,0.38246337890625004
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,1,power_law_1.01,0.39572608947753907
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,1,power_law_1.01,0.4194367980957031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,1,power_law_1.01,0.45672576904296874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,1,power_law_1.01,0.5292512130737305
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,1,power_law_1.01,0.6711199951171876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,1,power_law_1.01,0.8600672149658204
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,1,power_law_1.01,1.1171295928955078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,1,power_law_1.01,1.47759033203125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,1,power_law_1.01,2.2046885681152344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,1,power_law_1.01,2.898564453125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,1,power_law_1.01,4.268531799316406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,1,power_law_1.01,5.650557861328125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.2,1.2065894317626953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,1,power_law_1.01,9.050712890625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,1,power_law_1.01,0.06475903987884521
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,1,power_law_1.01,0.08255743980407715
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,1,power_law_1.01,0.1021235179901123
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,1,power_law_1.01,0.13600000381469726
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,1,power_law_1.01,0.147128324508667
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,1,power_law_1.01,0.14871423721313476
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,1,power_law_1.01,0.15002623558044434
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,1,power_law_1.01,0.14946559906005857
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,1,power_law_1.01,0.1509055995941162
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,1,power_law_1.01,0.1530419158935547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,1,power_law_1.01,0.1547961616516113
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,1,power_law_1.01,0.15594112396240234
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,1,power_law_1.01,0.15869248390197754
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,1,power_law_1.01,0.17275455474853516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,1,power_law_1.01,0.17894655227661133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,1,power_law_1.01,0.1812940788269043
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,1,power_law_1.01,0.19466751098632812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,1,power_law_1.01,0.23027263641357423
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,1,power_law_1.01,0.27596672058105465
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,1,power_law_1.01,0.3555904006958008
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,1,power_law_1.01,12.171290283203124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,1,power_law_1.01,0.4204070281982422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,1,power_law_1.01,0.5802163314819336
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,1,power_law_1.01,0.7661702728271484
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,1,power_law_1.01,1.1575212860107422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,4,1,power_law_1.2,1.4740371704101562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,1,power_law_1.01,1.5305305480957032
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,4,1,power_law_1.01,16.217768554687503
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,1,power_law_1.01,2.3265606689453127
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,1,power_law_1.01,3.327576904296875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,4,1,power_law_1.01,0.03209728002548218
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,4,1,power_law_1.01,0.07605055809020997
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,4,1,power_law_1.01,0.07808320045471191
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,4,1,power_law_1.01,0.07882304191589355
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,4,1,power_law_1.01,0.07874048233032227
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,4,1,power_law_1.01,0.07743103981018067
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,4,1,power_law_1.01,0.07725759983062744
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,4,1,power_law_1.01,0.07956480026245118
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,4,1,power_law_1.01,0.0800870418548584
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,4,1,power_law_1.01,4.126243286132812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,4,1,power_law_1.01,0.047945599555969234
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,4,1,power_law_1.01,0.056270718574523926
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,4,1,power_law_1.01,0.08032704353332519
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,4,1,power_law_1.01,0.0816140842437744
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,4,1,power_law_1.01,0.09065600395202636
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,4,1,power_law_1.01,0.08791551589965821
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,4,1,power_law_1.01,0.09400704383850098
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,4,1,power_law_1.01,0.10362688064575196
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,4,1,power_law_1.01,0.13013824462890625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.01,0.15289152145385743
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.01,0.20622079849243163
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.01,0.2542188835144043
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.01,0.3547903823852539
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.01,0.6665682983398438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.01,0.45727680206298826
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.01,0.8853625488281249
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,1,power_law_1.2,0.10353792190551758
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,1,power_law_1.2,0.12749823570251465
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,1,power_law_1.2,0.14084671974182128
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,1,power_law_1.2,0.23997823715209962
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,1,power_law_1.2,0.24071807861328126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,1,power_law_1.2,0.2507539176940918
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,1,power_law_1.2,0.25306943893432615
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,1,power_law_1.2,0.24898752212524414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,1,power_law_1.2,0.25968576431274415
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,1,power_law_1.2,0.2834329605102539
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,1,power_law_1.2,0.3461695861816406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,1,power_law_1.2,0.373691520690918
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,1,power_law_1.2,0.3788339233398438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,1,power_law_1.2,0.3954822540283203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,1,power_law_1.2,0.422685432434082
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,1,power_law_1.2,0.470055046081543
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,1,power_law_1.2,0.5245126342773437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,1,power_law_1.2,0.6752352142333985
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,1,power_law_1.2,0.8507091522216796
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.01,1.3510719299316407
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,1,power_law_1.2,1.1180633544921874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,1,power_law_1.2,1.506695098876953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,1,power_law_1.2,2.150880584716797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,1,power_law_1.2,3.012549743652344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.01,1.889024658203125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,1,power_law_1.2,4.246029357910156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,4,1,power_law_1.01,2.2582611083984374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,1,power_law_1.2,5.84523193359375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,1,power_law_1.2,0.06455423831939697
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,1,power_law_1.2,0.08053888320922852
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,1,power_law_1.2,0.08875776290893554
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,1,power_law_1.2,0.13682559967041014
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,1,power_law_1.2,0.14646400451660155
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,1,power_law_1.2,0.14893247604370116
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,1,power_law_1.2,0.15142911911010742
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,1,power_law_1.2,0.1503007984161377
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,1,power_law_1.2,0.1528819179534912
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,1,power_law_1.2,0.15390848159790038
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,1,power_law_1.2,0.15537792205810547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,1,power_law_1.2,0.1576518440246582
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,1,power_law_1.2,0.15796095848083497
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,1,power_law_1.2,0.173175048828125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,1,power_law_1.2,0.18264511108398437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,1,power_law_1.2,0.18432319641113282
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,1,power_law_1.2,0.19980607986450197
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,1,power_law_1.2,0.22972223281860354
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,1,power_law_1.2,0.2749254417419434
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,1,power_law_1.2,0.34199871063232423
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,1,power_law_1.2,0.41911041259765625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,1,power_law_1.2,9.656181640625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,1,power_law_1.2,0.5981087875366211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,1,power_law_1.2,0.7378412628173828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,1,power_law_1.2,1.1610546875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,1,power_law_1.2,12.141636962890626
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,1,power_law_1.2,1.5084812927246094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,4,1,power_law_1.2,15.415834960937499
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,4,1,power_law_1.2,0.03154815912246704
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,4,1,power_law_1.2,0.04623871803283691
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,4,1,power_law_1.2,0.05075647830963135
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,1,power_law_1.2,2.5063194274902343
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,4,1,power_law_1.2,0.07641088008880616
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,4,1,power_law_1.2,0.07452159881591797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,4,1,power_law_1.2,0.07885568141937256
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,4,1,power_law_1.2,0.07809408187866211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,4,1,power_law_1.2,0.07861120223999023
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,4,1,power_law_1.2,0.07707583904266357
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,4,1,power_law_1.2,0.0792953586578369
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,4,1,power_law_1.2,0.08004223823547363
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,4,1,power_law_1.2,0.08066752433776855
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,4,1,power_law_1.2,0.08820096015930176
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,1,power_law_1.2,3.229020080566406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,4,1,power_law_1.2,0.08082752227783203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,4,1,power_law_1.2,0.09362175941467285
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,4,1,power_law_1.2,0.09615039825439453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,4,1,power_law_1.2,0.10811584472656251
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,4,1,power_law_1.2,4.318856811523437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,4,1,power_law_1.2,0.13214336395263673
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.2,0.15466752052307128
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.2,0.20475584030151367
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.2,0.2596883201599121
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.2,0.36580352783203124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.2,0.4682841491699219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.2,0.6893299102783204
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.2,0.8946918487548828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,1,power_law_1.01,0.08146816253662109
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,1,power_law_1.01,0.10167936325073242
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,1,power_law_1.01,0.14420031547546386
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,1,power_law_1.01,0.2853350448608398
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,1,power_law_1.01,0.39104064941406247
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,1,power_law_1.01,0.5131718444824218
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,1,power_law_1.01,0.6927993774414063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,1,power_law_1.01,0.7237689971923829
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,1,power_law_1.01,0.7351974487304688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,1,power_law_1.01,0.7439347076416015
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,1,power_law_1.01,0.7819884490966797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,1,power_law_1.01,0.8380051422119141
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,1,power_law_1.01,0.872442855834961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,1,power_law_1.01,0.9267481231689454
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,1,power_law_1.01,0.9796281433105468
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,1,power_law_1.01,0.988076171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,1,power_law_1.01,1.0180358123779296
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,1,power_law_1.01,1.1333612823486328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,1,power_law_1.01,1.2255340576171876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.2,1.3652806091308594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,1,power_law_1.01,1.4539456176757812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,1,power_law_1.01,1.5995864868164062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,1,power_law_1.01,2.0814361572265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,1,power_law_1.01,2.513427124023438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,1,power_law_1.01,3.6616256713867186
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,1,power_law_1.01,4.216795654296876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,1,power_law_1.01,6.315651245117188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,1,power_law_1.01,7.862139892578125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,1,power_law_1.01,0.05298240184783935
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,1,power_law_1.01,0.07356800079345703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.2,1.8600492858886717
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,1,power_law_1.01,0.10501248359680175
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,1,power_law_1.01,0.25534400939941404
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,1,power_law_1.01,0.3505465698242188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,1,power_law_1.01,0.4403302383422852
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,1,power_law_1.01,0.46885822296142576
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,4,1,power_law_1.01,10.7121826171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,1,power_law_1.01,0.4776236724853516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,1,power_law_1.01,0.48525310516357417
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,1,power_law_1.01,0.5013401412963867
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,1,power_law_1.01,0.5167180633544921
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,1,power_law_1.01,0.17447807312011718
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,1,power_law_1.01,0.5202764892578126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,1,power_law_1.01,0.5325491333007812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,1,power_law_1.01,0.591319694519043
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,1,power_law_1.01,0.5925657653808594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,1,power_law_1.01,0.623520622253418
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,1,power_law_1.01,0.6537728118896484
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,1,power_law_1.01,0.6837484741210937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,1,power_law_1.01,0.7660269165039063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,1,power_law_1.01,0.8372473907470702
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,1,power_law_1.01,0.9775251007080078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,1,power_law_1.01,1.1751967620849608
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,1,power_law_1.01,1.4822431945800782
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,1,power_law_1.01,1.8064607238769532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,1,power_law_1.01,2.59399169921875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,1,power_law_1.01,3.1155206298828126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,4,1,power_law_1.01,3.8521582031249997
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,4,1,power_law_1.2,2.3323667907714842
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,1,power_law_1.01,0.06282112121582031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,1,power_law_1.01,0.0917632007598877
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,1,power_law_1.01,0.16166528701782226
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,1,power_law_1.01,0.07635456085205078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,1,power_law_1.01,0.19730688095092774
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,1,power_law_1.01,0.27114688873291015
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,1,power_law_1.01,0.1221440029144287
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,1,power_law_1.01,0.2839321517944336
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,1,power_law_1.01,0.2941056060791015
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,1,power_law_1.01,0.29828224182128904
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,1,power_law_1.01,0.3134297561645508
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,1,power_law_1.01,0.3231148910522461
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,1,power_law_1.01,0.3351532745361328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,1,power_law_1.01,0.34359169006347656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,1,power_law_1.01,0.3587200164794922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,1,power_law_1.01,0.36819583892822266
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,1,power_law_1.01,0.3961401748657226
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,1,power_law_1.01,0.4450796890258789
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.01,0.7109728240966797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.01,0.4973651123046875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.01,0.6063500976562499
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.01,1.1772838592529298
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.01,0.9177779388427734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.01,1.628137664794922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.01,2.171613464355469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,1,power_law_1.2,0.08156352043151856
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,1,power_law_1.2,0.10254400253295899
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,1,power_law_1.2,0.1441919994354248
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,1,power_law_1.2,0.25543872833251957
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,1,power_law_1.2,0.36453121185302734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,1,power_law_1.2,0.4907622528076172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,1,power_law_1.2,0.6777113342285157
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,1,power_law_1.2,0.70714111328125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,1,power_law_1.2,0.7198687744140625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,1,power_law_1.2,0.7418937683105469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,1,power_law_1.2,0.7806265258789062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,1,power_law_1.2,0.8301766204833985
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,1,power_law_1.2,0.8688172912597656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,1,power_law_1.2,0.9504691314697264
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,4,1,power_law_1.01,3.184523620605469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,4,1,power_law_1.01,4.211903991699218
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,1,power_law_1.2,0.979135971069336
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,1,power_law_1.2,1.0029593658447264
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,7168,2048,8,256,4,1,power_law_1.01,5.1692614746093755
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,1,power_law_1.2,1.1686265563964844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,1,power_law_1.2,1.2393708801269532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,1,power_law_1.2,1.5268435668945313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,1,power_law_1.2,1.0439167785644532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,1,power_law_1.2,1.7086688232421874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,1,power_law_1.2,2.309534149169922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,1,power_law_1.2,2.781603088378906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,1,power_law_1.2,3.420133666992187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,1,power_law_1.2,4.259086608886719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,1,power_law_1.2,0.051029119491577145
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,1,power_law_1.2,0.07312320232391357
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,1,power_law_1.2,0.1054304027557373
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,1,power_law_1.2,0.16528448104858398
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,1,power_law_1.2,6.694464111328125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,1,power_law_1.2,0.24395328521728515
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,1,power_law_1.2,0.3374220657348633
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,1,power_law_1.2,0.4337171173095703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,1,power_law_1.2,0.45505729675292966
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,1,power_law_1.2,0.4707417678833008
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,1,power_law_1.2,0.48825664520263673
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,1,power_law_1.2,8.505840454101563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,1,power_law_1.2,0.49607936859130863
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,1,power_law_1.2,0.513175048828125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,1,power_law_1.2,0.5195315170288086
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,1,power_law_1.2,0.540810890197754
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,1,power_law_1.2,0.5981203079223633
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,1,power_law_1.2,0.612784652709961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,1,power_law_1.2,0.6314835357666015
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,4,1,power_law_1.2,9.844254150390626
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,1,power_law_1.2,0.6793772888183593
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,1,power_law_1.2,0.7111392211914063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,1,power_law_1.2,0.7899616241455079
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,1,power_law_1.2,0.8660409545898438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,1,power_law_1.2,1.0443122863769532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,1,power_law_1.2,1.2000070190429688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,1,power_law_1.2,1.5023794555664063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,1,power_law_1.2,1.8593312072753907
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,1,power_law_1.2,2.5551104736328125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,1,power_law_1.2,0.06328256130218506
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,1,power_law_1.2,3.2069952392578123
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,1,power_law_1.2,0.07684864044189453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,1,power_law_1.2,0.08844287872314453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,1,power_law_1.2,0.11629695892333984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,4,1,power_law_1.2,3.7575399780273435
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,1,power_law_1.2,0.15125375747680664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,1,power_law_1.2,0.18871871948242186
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,1,power_law_1.2,0.26653440475463863
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,1,power_law_1.2,0.2900601577758789
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,1,power_law_1.2,0.27707456588745116
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,1,power_law_1.2,0.3012985610961914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,1,power_law_1.2,0.31332544326782225
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,1,power_law_1.2,0.3254502487182617
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,1,power_law_1.2,0.3354822540283203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,1,power_law_1.2,0.35747329711914066
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,1,power_law_1.2,0.3769657516479492
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,1,power_law_1.2,0.3883411026000977
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,1,power_law_1.2,0.4313011169433594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,1,power_law_1.2,0.5064742279052734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.2,0.5838399887084961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.2,0.9411424255371094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.2,0.7422898864746094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.2,1.3111494445800782
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.2,1.7040269470214846
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.2,2.469402313232422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,1,power_law_1.01,0.0288755202293396
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,1,power_law_1.01,0.044188799858093264
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,1,power_law_1.01,0.05097599983215332
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,1,power_law_1.01,0.05907328128814697
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,1,power_law_1.01,0.0775430393218994
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,1,power_law_1.01,0.07759424209594726
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,1,power_law_1.01,0.07937088012695312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,1,power_law_1.01,0.08247936248779297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,1,power_law_1.01,0.08377023696899413
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,1,power_law_1.01,0.0897708797454834
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,1,power_law_1.01,0.09660415649414063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,1,power_law_1.01,0.09814784049987793
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,1,power_law_1.01,0.10154879570007325
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,1,power_law_1.01,0.10568767547607423
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,1,power_law_1.01,0.10795583724975585
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.2,3.361679992675781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,1,power_law_1.01,0.11505727767944336
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,1,power_law_1.01,0.1352403163909912
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,1,power_law_1.01,0.15617664337158205
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,1,power_law_1.01,0.18307199478149414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,1,power_law_1.01,0.21556224822998046
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,1,power_law_1.01,0.27938816070556644
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,1,power_law_1.01,0.3496480178833008
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,1,power_law_1.01,0.4778688049316406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,1,power_law_1.01,0.6163558578491211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,1,power_law_1.01,0.8784505462646484
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,1,power_law_1.01,1.1481721496582031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2048,768,8,128,4,1,power_law_1.01,1.3985848999023438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,1,power_law_1.01,0.03331007957458496
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,1,power_law_1.01,0.039882879257202145
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,1,power_law_1.01,0.03173631906509399
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,1,power_law_1.01,0.048024959564208984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,1,power_law_1.01,0.07467584133148193
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,1,power_law_1.01,0.07914944171905518
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,1,power_law_1.01,0.07984447956085206
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,1,power_law_1.01,0.081976318359375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,1,power_law_1.01,0.057368321418762205
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,1,power_law_1.01,0.08196672439575195
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,1,power_law_1.01,0.0958944034576416
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,1,power_law_1.01,0.09717887878417969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,1,power_law_1.01,0.0989791965484619
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,1,power_law_1.01,0.10075391769409178
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,1,power_law_1.01,0.10294015884399414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,1,power_law_1.01,0.08422911643981934
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,1,power_law_1.01,0.10971199989318847
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,1,power_law_1.01,0.1163967990875244
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,1,power_law_1.01,0.12992704391479493
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,1,power_law_1.01,0.15254207611083986
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,1,power_law_1.01,0.1770751953125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,1,power_law_1.01,0.22348287582397458
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,1,power_law_1.01,0.26966463088989256
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,1,power_law_1.01,0.36959808349609374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,1,power_law_1.01,0.4599750518798828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,1,power_law_1.01,0.6644671630859376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,1,power_law_1.01,0.84953857421875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,2048,768,8,128,4,1,power_law_1.01,1.0461325073242187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,1,power_law_1.01,0.04158592224121094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,1,power_law_1.01,0.04452991962432861
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,1,power_law_1.01,0.04735680103302002
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,1,power_law_1.01,0.04027584075927734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,1,power_law_1.01,0.05061823844909667
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,1,power_law_1.01,0.05894720077514648
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,1,power_law_1.01,0.06141568183898925
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,1,power_law_1.01,0.06185408115386963
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,1,power_law_1.01,0.06329599857330323
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,1,power_law_1.01,0.06341887950897217
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,1,power_law_1.01,0.06508800029754638
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,1,power_law_1.01,0.06597504138946533
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,1,power_law_1.01,0.06778560161590577
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,1,power_law_1.01,0.07030848026275635
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,1,power_law_1.01,0.07387263774871826
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,1,power_law_1.01,0.07512767791748047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,1,power_law_1.01,0.09185919761657715
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,1,power_law_1.01,0.08041728019714356
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,1,power_law_1.01,0.1078656005859375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,1,power_law_1.01,0.13455615997314455
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,1,power_law_1.01,0.1630284881591797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,1,power_law_1.01,0.22515520095825195
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,1,power_law_1.01,0.28905664443969725
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,1,power_law_1.01,0.42200897216796873
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,1,power_law_1.01,0.554791030883789
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,1,power_law_1.01,0.8404243469238282
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,1,power_law_1.01,1.1402003479003906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,1,power_law_1.2,0.04281599998474121
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,1,power_law_1.2,0.02924799919128418
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,1,power_law_1.2,0.057566080093383785
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,1,power_law_1.2,0.03721856117248536
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,1,power_law_1.2,0.07925695896148681
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,1,power_law_1.2,0.07806911945343017
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,1,power_law_1.2,0.08421567916870118
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,1,power_law_1.2,0.08226176261901855
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,1,power_law_1.2,0.0857817554473877
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,1,power_law_1.2,0.09002047538757324
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,1,power_law_1.2,0.09720895767211915
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,1,power_law_1.2,0.049132161140441895
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,1,power_law_1.2,0.09900799751281739
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,1,power_law_1.2,0.10311807632446288
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,1,power_law_1.2,0.10815744400024414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,1,power_law_1.2,0.11479359626770021
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,1,power_law_1.2,0.13809920310974122
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,1,power_law_1.2,0.15691519737243653
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,1,power_law_1.2,0.1851161575317383
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,1,power_law_1.2,0.10416383743286133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,1,power_law_1.2,0.21659904479980469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,1,power_law_1.2,0.2820684814453125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,1,power_law_1.2,0.34759166717529294
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2048,768,8,128,4,1,power_law_1.01,1.4338272094726563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,1,power_law_1.2,0.48528255462646486
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,1,power_law_1.2,0.6178976058959961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,1,power_law_1.2,0.03253119945526123
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,1,power_law_1.2,0.034529919624328616
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,1,power_law_1.2,0.04148287773132324
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,1,power_law_1.2,0.048024959564208984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,1,power_law_1.2,0.05690176010131835
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,1,power_law_1.2,0.07801792144775391
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,1,power_law_1.2,0.0802073574066162
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,1,power_law_1.2,0.08203712463378907
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,1,power_law_1.2,0.08369407653808594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,1,power_law_1.2,0.08394432067871094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,1,power_law_1.2,0.08413375854492187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,1,power_law_1.2,0.0963219165802002
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,1,power_law_1.2,0.09763135910034179
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,1,power_law_1.2,0.10037952423095704
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,1,power_law_1.2,0.1030297565460205
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,1,power_law_1.2,0.10614656448364257
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,1,power_law_1.2,0.11102335929870606
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,1,power_law_1.2,0.1186086368560791
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,1,power_law_1.2,0.13586432456970216
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,1,power_law_1.2,0.1575475215911865
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,1,power_law_1.2,0.18126720428466797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,1,power_law_1.2,0.22458112716674802
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,1,power_law_1.2,0.27517183303833004
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,1,power_law_1.2,0.8828710174560547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,1,power_law_1.2,0.3721548843383789
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,1,power_law_1.2,0.46244224548339846
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,1,power_law_1.2,1.1445894622802735
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,1,power_law_1.2,0.6657222747802735
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,1,power_law_1.2,0.04039423942565918
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,1,power_law_1.2,0.04195072174072266
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,1,power_law_1.2,0.04372416019439697
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,1,power_law_1.2,0.046394882202148435
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,1,power_law_1.2,0.049696640968322756
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,1,power_law_1.2,0.0594540786743164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,1,power_law_1.2,0.06162303924560547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,1,power_law_1.2,0.06315199851989746
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,1,power_law_1.2,0.0639027214050293
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,1,power_law_1.2,0.06472576141357422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,1,power_law_1.2,0.0657094383239746
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,1,power_law_1.2,0.06812799930572509
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,1,power_law_1.2,0.06903999805450439
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,1,power_law_1.2,0.07310143947601319
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,1,power_law_1.2,0.07660160064697266
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,1,power_law_1.2,0.0807801628112793
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2048,768,8,128,4,1,power_law_1.2,1.417211456298828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,1,power_law_1.2,0.08892416000366211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,1,power_law_1.2,0.8534400177001953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,1,power_law_1.2,0.10832127571105957
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,1,power_law_1.2,0.12969216346740722
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,1,power_law_1.2,0.17405311584472655
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,1,power_law_1.2,0.2216985511779785
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,1,power_law_1.2,0.3201612854003906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,1,power_law_1.2,0.4211052703857422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,1,power_law_1.2,0.6259276962280274
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,2048,768,8,128,4,1,power_law_1.2,1.0408646392822267
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,1,power_law_1.2,0.8212159729003906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,1,power_law_1.01,0.03907392024993896
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,1,power_law_1.01,0.06445375919342042
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,1,power_law_1.01,0.0834284782409668
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,1,power_law_1.01,0.11054400444030761
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,1,power_law_1.01,0.13705599784851075
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,1,power_law_1.01,0.22238143920898437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,1,power_law_1.01,0.22516927719116211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,1,power_law_1.01,0.22939327239990231
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,1,power_law_1.01,0.23294015884399416
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,1,power_law_1.01,0.23009471893310546
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,1,power_law_1.01,0.24485696792602538
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,1,power_law_1.01,0.2637759971618652
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,1,power_law_1.01,0.266888313293457
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,1,power_law_1.01,0.2725279998779297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,1,power_law_1.01,0.2773504066467285
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,1,power_law_1.01,0.28189888000488283
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,1,power_law_1.01,0.29535104751586916
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,1,power_law_1.01,0.36235393524169923
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,1,power_law_1.01,0.3912300872802734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,1,power_law_1.01,0.47600128173828127
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,1,power_law_1.01,0.5651839828491211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,1,power_law_1.01,0.7454252624511719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,1,power_law_1.01,0.9226464080810548
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,1,power_law_1.01,1.2884921264648437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,1,power_law_1.2,1.2262694549560547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,1,power_law_1.01,1.6747596740722657
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,1,power_law_1.01,2.417689666748047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,1,power_law_1.01,3.1832116699218753
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,1,power_law_1.01,0.03777791976928711
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,1,power_law_1.01,0.050264320373535155
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,1,power_law_1.01,0.06809792041778565
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,1,power_law_1.01,0.08931391716003419
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,1,power_law_1.01,0.11771391868591308
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,1,power_law_1.01,0.1555891227722168
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,4,1,power_law_1.01,4.0319091796875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,1,power_law_1.01,0.1587775993347168
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,1,power_law_1.01,0.16086591720581053
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,1,power_law_1.01,0.1641139221191406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,1,power_law_1.01,0.16563007354736328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,1,power_law_1.01,0.16938495635986328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,1,power_law_1.01,0.1887308883666992
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,1,power_law_1.01,0.19192384719848632
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,1,power_law_1.01,0.19458303451538086
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,1,power_law_1.01,0.19985343933105468
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,1,power_law_1.01,0.20058687210083007
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,1,power_law_1.01,0.21304191589355467
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,1,power_law_1.01,0.22541248321533205
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,1,power_law_1.01,0.2539558410644531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,1,power_law_1.01,0.2865555191040039
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,1,power_law_1.01,0.3442451095581055
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,1,power_law_1.01,0.4279808044433594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,1,power_law_1.01,0.5154489517211914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,1,power_law_1.01,0.7018482971191407
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,1,power_law_1.01,0.8823654174804687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,1,power_law_1.01,1.2415539550781252
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,1,power_law_1.01,1.607737579345703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,4,1,power_law_1.01,1.954010314941406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,1,power_law_1.01,0.039839999675750734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,1,power_law_1.01,0.053013758659362795
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,1,power_law_1.01,0.04434815883636475
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,1,power_law_1.01,0.06392064094543456
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,1,power_law_1.01,0.07228928089141846
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,1,power_law_1.01,0.09502719879150391
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,1,power_law_1.01,0.10000127792358397
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,1,power_law_1.2,1.6376039123535158
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,1,power_law_1.01,0.10217408180236816
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,1,power_law_1.01,0.10461183547973633
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,1,power_law_1.01,0.11062399864196777
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,1,power_law_1.01,0.11319295883178711
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,1,power_law_1.01,0.11620160102844239
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,1,power_law_1.01,0.12236031532287597
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,1,power_law_1.01,0.106428804397583
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,1,power_law_1.01,0.13139455795288085
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,1,power_law_1.01,0.1358521556854248
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,1,power_law_1.01,0.1522982406616211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,1,power_law_1.01,0.17918720245361328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.01,0.21096448898315429
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2048,768,8,128,4,1,power_law_1.2,2.0476332092285157
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.01,0.2767897605895996
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.01,0.3464761734008789
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.01,0.48595390319824217
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.01,0.6343040084838867
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.01,0.9432230377197266
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.01,1.2676985931396483
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,1,power_law_1.2,0.04750912189483643
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,1,power_law_1.2,0.0661574411392212
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,1,power_law_1.2,0.08274880409240723
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.01,1.9021420288085937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,1,power_law_1.2,0.10450495719909667
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,1,power_law_1.2,0.22387775421142578
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,1,power_law_1.2,0.229880313873291
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,1,power_law_1.2,0.12652159690856934
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,1,power_law_1.2,0.23314176559448244
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,1,power_law_1.2,0.2305740737915039
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,1,power_law_1.2,0.2354521560668945
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,1,power_law_1.2,0.24545536041259766
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,1,power_law_1.2,0.2643020820617676
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,1,power_law_1.2,0.26894720077514644
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,1,power_law_1.2,0.27458431243896486
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,1,power_law_1.2,0.27862911224365233
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,1,power_law_1.2,0.28290496826171874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,1,power_law_1.2,0.30362367630004883
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,1,power_law_1.2,0.36744190216064454
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,1,power_law_1.2,0.3926988983154297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,1,power_law_1.2,0.48974655151367186
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,1,power_law_1.2,0.563699836730957
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,1,power_law_1.2,0.7474918365478516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,1,power_law_1.2,0.9320089721679687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,1,power_law_1.2,1.2999565124511718
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,4,1,power_law_1.01,3.217405395507812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.01,2.5416761779785153
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,1,power_law_1.2,1.7275462341308594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,1,power_law_1.2,0.037880959510803225
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,1,power_law_1.2,0.04870207786560059
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,1,power_law_1.2,0.06470143795013428
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,1,power_law_1.2,0.07926911830902099
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,1,power_law_1.2,0.10904319763183594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,1,power_law_1.2,0.1499116802215576
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,1,power_law_1.2,0.15877375602722169
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,1,power_law_1.2,0.1646931266784668
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,1,power_law_1.2,0.16818944931030272
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,1,power_law_1.2,0.1683020782470703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,1,power_law_1.2,0.17165184020996094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,1,power_law_1.2,0.19452991485595703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,1,power_law_1.2,0.19522623062133787
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,1,power_law_1.2,0.20200832366943358
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,1,power_law_1.2,0.20458240509033204
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,1,power_law_1.2,0.20868928909301757
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,1,power_law_1.2,0.2141209602355957
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,1,power_law_1.2,0.23245887756347655
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,1,power_law_1.2,0.2573075294494629
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,1,power_law_1.2,0.2996492767333984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,1,power_law_1.2,0.3447622299194336
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,1,power_law_1.2,0.4313049697875977
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,1,power_law_1.2,0.5210054397583008
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,1,power_law_1.2,2.3831756591796873
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,1,power_law_1.2,0.706170883178711
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,1,power_law_1.2,0.884246368408203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,1,power_law_1.2,3.2330859375000003
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,1,power_law_1.2,1.2456102752685547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,1,power_law_1.2,0.03982719898223877
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,1,power_law_1.2,0.04456128120422363
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,1,power_law_1.2,0.05230783939361572
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,1,power_law_1.2,0.059784960746765134
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,1,power_law_1.2,0.06983935832977295
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,1,power_law_1.2,0.0955737590789795
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,1,power_law_1.2,0.10057087898254395
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,1,power_law_1.2,0.1029631996154785
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,1,power_law_1.2,0.10575103759765625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,1,power_law_1.2,0.1075705623626709
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,1,power_law_1.2,0.112042875289917
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,1,power_law_1.2,0.11658047676086425
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,1,power_law_1.2,0.11981375694274903
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,1,power_law_1.2,0.12961215972900392
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,1,power_law_1.2,0.13921471595764162
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,1,power_law_1.2,0.14847488403320314
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,1,power_law_1.2,0.1689753532409668
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,1,power_law_1.2,0.21023359298706054
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,1,power_law_1.2,1.609752960205078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,4,1,power_law_1.2,3.8317413330078125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.2,0.25616960525512694
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.2,0.3550899124145508
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.2,0.45306049346923827
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.2,0.6696761322021485
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.2,0.9009702301025391
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,4,1,power_law_1.2,1.969917449951172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.2,1.3315097045898436
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.2,1.765776672363281
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,1,power_law_1.01,0.0810694408416748
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,1,power_law_1.01,0.1021299171447754
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,1,power_law_1.01,0.1530016040802002
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,1,power_law_1.01,0.2576684761047363
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,1,power_law_1.01,0.33203136444091796
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,1,power_law_1.01,0.4956819152832031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,1,power_law_1.01,0.5205696105957032
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,1,power_law_1.01,0.5466144180297852
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,1,power_law_1.01,0.571325454711914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,1,power_law_1.01,0.5828511810302734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,1,power_law_1.01,0.6114988708496094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,1,power_law_1.01,0.6565074920654297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,1,power_law_1.01,0.6727884674072266
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,1,power_law_1.01,0.7343762969970703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,1,power_law_1.01,0.7536319732666016
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,1,power_law_1.01,0.762903060913086
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,1,power_law_1.01,0.7949196624755859
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,1,power_law_1.01,0.8840505981445312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,1,power_law_1.01,0.989780502319336
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,1,power_law_1.01,1.1605907440185548
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,1,power_law_1.01,1.4671974182128906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,1,power_law_1.01,1.9500274658203125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,1,power_law_1.01,2.4328953552246095
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.2,2.6218612670898436
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,1,power_law_1.01,3.3984307861328125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,1,power_law_1.01,4.004807739257812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,1,power_law_1.01,6.227196044921875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,1,power_law_1.01,8.001502075195313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,1,power_law_1.01,0.05456960201263428
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,1,power_law_1.01,0.07565951824188233
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,1,power_law_1.01,0.12102208137512208
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,1,power_law_1.01,0.15952768325805664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,1,power_law_1.01,0.22096511840820315
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,1,power_law_1.01,0.32853694915771486
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,1,power_law_1.01,0.33358718872070314
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,1,power_law_1.01,0.34213504791259763
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,4,1,power_law_1.01,10.674451904296875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,1,power_law_1.01,0.35014591217041013
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,1,power_law_1.01,0.3526547241210937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,1,power_law_1.01,0.3622963333129883
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,1,power_law_1.01,0.37199680328369145
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,1,power_law_1.01,0.41359294891357423
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,1,power_law_1.01,0.4286342239379882
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,1,power_law_1.01,0.4337542343139648
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,1,power_law_1.01,0.42050815582275386
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,1,power_law_1.01,0.4583443069458008
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,1,power_law_1.01,0.48238273620605465
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,1,power_law_1.01,0.5206918334960937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,1,power_law_1.01,0.5836435317993164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,1,power_law_1.01,0.6831769561767578
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,1,power_law_1.01,0.8345753479003907
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,1,power_law_1.01,1.0021126556396485
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.2,3.4961056518554683
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,1,power_law_1.01,1.3587513732910157
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,1,power_law_1.01,1.7315225219726564
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,1,power_law_1.01,2.338564453125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,1,power_law_1.01,2.9664633178710935
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,4,1,power_law_1.01,3.7085287475585935
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,1,power_law_1.01,0.047461757659912104
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,1,power_law_1.01,0.06562047958374023
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,1,power_law_1.01,0.101528959274292
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,1,power_law_1.01,0.12668543815612793
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,1,power_law_1.01,0.1887436866760254
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,1,power_law_1.01,0.206625919342041
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,1,power_law_1.01,0.19848768234252928
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,1,power_law_1.01,0.079137282371521
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,1,power_law_1.01,0.21330240249633786
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,1,power_law_1.01,0.21560768127441404
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,1,power_law_1.01,0.22431488037109376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,1,power_law_1.01,0.23145408630371095
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,1,power_law_1.01,0.2347808074951172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,1,power_law_1.01,0.24938880920410159
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,1,power_law_1.01,0.25730560302734373
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,1,power_law_1.01,0.2618720054626465
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.01,0.3826860809326172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,1,power_law_1.01,0.29084287643432616
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,1,power_law_1.01,0.33794689178466797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,4,1,power_law_1.2,4.350570373535156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.01,0.4815161514282226
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.01,0.5959334564208985
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.01,0.8125260925292969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.01,1.035823974609375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.01,1.508153533935547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.01,1.9902316284179686
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.01,2.9556781005859376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,1,power_law_1.2,0.10134528160095216
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,1,power_law_1.2,0.15134655952453613
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,1,power_law_1.2,0.23201343536376956
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,1,power_law_1.2,0.08119359970092774
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,1,power_law_1.2,0.3182931137084961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,1,power_law_1.2,0.5118675231933594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,1,power_law_1.2,0.5029734420776367
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,1,power_law_1.2,0.547470703125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,1,power_law_1.2,0.5816409683227539
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,1,power_law_1.2,0.5998150253295899
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,1,power_law_1.2,0.6355142211914062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,1,power_law_1.2,0.6774988555908203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,1,power_law_1.2,0.7154796600341797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,1,power_law_1.2,0.7512786865234375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,1,power_law_1.2,0.7695500946044922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,1,power_law_1.2,0.7790348815917969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,1,power_law_1.2,0.810980453491211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,1,power_law_1.2,0.9020230102539063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,1,power_law_1.2,0.9709580993652344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,1,power_law_1.2,1.2052499389648437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,1,power_law_1.2,1.5100531005859374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,1,power_law_1.2,1.9481272888183594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,1,power_law_1.2,2.4453305053710936
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,1,power_law_1.2,3.3909460449218747
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,1,power_law_1.2,4.189820861816406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.01,3.953795776367188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,2560,8,160,4,1,power_law_1.01,4.9665856933593755
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,1,power_law_1.2,6.166727294921875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,1,power_law_1.2,0.07384448051452637
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,1,power_law_1.2,0.06070271968841553
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,1,power_law_1.2,0.11251711845397949
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,1,power_law_1.2,0.1498534393310547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,1,power_law_1.2,0.206046085357666
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,1,power_law_1.2,0.3231123352050781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,1,power_law_1.2,0.32915199279785157
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,1,power_law_1.2,0.3447795104980469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,1,power_law_1.2,0.3537696075439453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,1,power_law_1.2,0.35989505767822266
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,1,power_law_1.2,0.36905025482177733
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,1,power_law_1.2,0.37698368072509764
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,1,power_law_1.2,0.4288460922241211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,1,power_law_1.2,0.43593921661376955
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,1,power_law_1.2,0.4428140640258789
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,1,power_law_1.2,0.4552223968505859
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,1,power_law_1.2,0.4681856155395508
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,1,power_law_1.2,0.49171264648437496
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,1,power_law_1.2,0.5315609741210937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,1,power_law_1.2,0.6116313552856445
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,1,power_law_1.2,0.6882899475097657
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,1,power_law_1.2,0.8606221008300782
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,1,power_law_1.2,1.007609634399414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,1,power_law_1.2,1.3547608947753906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,1,power_law_1.2,1.7040229797363282
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,1,power_law_1.2,8.51118408203125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,1,power_law_1.2,2.3216876220703124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,1,power_law_1.2,0.04815807819366455
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,1,power_law_1.2,0.06261631965637207
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,1,power_law_1.2,0.07652607917785645
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,1,power_law_1.2,0.09569664001464843
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,1,power_law_1.2,0.12015423774719239
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,1,power_law_1.2,0.1860927963256836
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,1,power_law_1.2,0.19672832489013672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,1,power_law_1.2,0.20777471542358397
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,1,power_law_1.2,0.21644351959228514
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,1,power_law_1.2,3.0675604248046873
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,1,power_law_1.2,0.22179712295532225
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,1,power_law_1.2,0.23109952926635743
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,4,1,power_law_1.2,9.860914306640625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,1,power_law_1.2,0.2358220863342285
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,1,power_law_1.2,0.2581772804260254
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,4,1,power_law_1.2,3.7354925537109374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,1,power_law_1.2,0.27729856491088867
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,1,power_law_1.2,0.29029375076293945
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,1,power_law_1.2,0.38686912536621093
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,1,power_law_1.2,0.24650560379028322
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,1,power_law_1.2,0.32408958435058594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.2,0.45758911132812496
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.2,0.6163232040405273
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.2,0.7822866821289063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.2,1.152032012939453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.2,1.5388986206054687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.2,2.3458029174804684
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.2,3.167528381347656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,1,power_law_1.01,0.08547200202941894
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,1,power_law_1.01,0.10573439598083496
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,1,power_law_1.01,0.14744383811950684
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,1,power_law_1.01,0.29070783615112306
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,1,power_law_1.01,0.47467071533203126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,1,power_law_1.01,0.6349811172485351
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,1,power_law_1.01,0.7314406585693359
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.2,4.856136169433594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,1,power_law_1.01,0.9218649291992188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,1,power_law_1.01,0.9697702026367188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,1,power_law_1.01,1.0040953826904295
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,1,power_law_1.01,1.063562240600586
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,1,power_law_1.01,1.0629837036132812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,1,power_law_1.01,1.1424928283691407
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,1,power_law_1.01,1.1886732482910156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,1,power_law_1.01,1.283272247314453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,1,power_law_1.01,1.3614508056640626
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,1,power_law_1.01,1.3744064331054688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,1,power_law_1.01,1.4537274169921877
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,1,power_law_1.01,1.5852403259277343
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,1,power_law_1.01,1.7795481872558594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,1,power_law_1.01,2.0736813354492187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,1,power_law_1.01,2.94052490234375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,1,power_law_1.01,4.154973449707031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,1,power_law_1.01,4.6736141967773435
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.2,6.509513549804687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,1,power_law_1.01,7.0735888671875005
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,1,power_law_1.01,0.05269951820373535
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,1,power_law_1.01,0.07694272041320802
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,1,power_law_1.01,0.10899519920349121
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,1,power_law_1.01,0.18072448730468751
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,1,power_law_1.01,8.489609375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,1,power_law_1.01,0.30808832168579103
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,1,power_law_1.01,0.42504703521728515
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,1,power_law_1.01,0.4989145660400391
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,1,power_law_1.01,0.6021145629882813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,1,power_law_1.01,0.6326892852783204
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,1,power_law_1.01,0.6569964599609375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,1,power_law_1.01,0.6766310119628907
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,1,power_law_1.01,0.6824678039550781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,1,power_law_1.01,0.7069843292236329
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,1,power_law_1.01,0.7248998260498046
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,4,1,power_law_1.01,10.606689453125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,1,power_law_1.01,0.7350514984130859
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,1,power_law_1.01,0.7568966674804687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,1,power_law_1.01,0.8573056030273438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,1,power_law_1.01,0.8685177612304689
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,1,power_law_1.01,0.9016563415527343
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,1,power_law_1.01,0.9836121368408204
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,1,power_law_1.01,1.078993911743164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,1,power_law_1.01,1.222644500732422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,1,power_law_1.01,1.426373748779297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,1,power_law_1.01,1.8155532836914063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,1,power_law_1.01,2.060832672119141
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,1,power_law_1.01,2.8097216796874998
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,1,power_law_1.01,3.5043936157226563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,4,1,power_law_1.01,4.108560791015625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,1,power_law_1.01,0.07806528091430665
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,1,power_law_1.01,0.07673088073730469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,1,power_law_1.01,0.08207167625427246
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,1,power_law_1.01,0.0858022403717041
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,1,power_law_1.01,0.0988492774963379
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,1,power_law_1.01,0.10629631996154784
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,1,power_law_1.01,0.11777791976928711
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,1,power_law_1.01,0.12529024124145507
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,1,power_law_1.01,0.07900479793548584
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,1,power_law_1.01,0.13779007911682128
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,1,power_law_1.01,0.15915776252746583
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,1,power_law_1.01,0.2219308853149414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,1,power_law_1.01,0.2679212760925293
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,1,power_law_1.01,0.19792896270751953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,1,power_law_1.01,0.3238585662841797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,1,power_law_1.01,0.3710649490356445
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,1,power_law_1.01,0.46768638610839847
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,1,power_law_1.01,0.8627986907958984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,1,power_law_1.01,0.6625856018066406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,1,power_law_1.01,1.3004672241210937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,1,power_law_1.01,1.7067430114746096
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,1,power_law_1.01,2.508477478027344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,1,power_law_1.01,3.3226739501953126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,1,power_law_1.2,0.09072511672973634
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,1,power_law_1.2,0.10470208168029785
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,1,power_law_1.2,0.14696384429931642
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,1,power_law_1.2,0.2900307273864746
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,1,power_law_1.2,0.4148569488525391
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,1,power_law_1.2,0.5807347106933594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,1,power_law_1.2,0.6821279907226563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,1,power_law_1.2,0.9058918762207032
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,1,power_law_1.2,0.9409216308593751
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,1,power_law_1.2,0.9668185424804687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,1,power_law_1.01,4.950159912109375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,1,power_law_1.2,1.0097382354736328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,1,power_law_1.2,1.074109420776367
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,1,power_law_1.2,1.1166893005371095
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,1,power_law_1.2,1.2103238677978516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,1,power_law_1.2,1.27728515625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,1,power_law_1.2,1.4216773986816407
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,1,power_law_1.2,1.3822221374511718
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,1,power_law_1.2,1.5484608459472657
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,1,power_law_1.2,1.5812524414062499
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,1,power_law_1.2,1.896321258544922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,1,power_law_1.2,2.1027789306640625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,1,power_law_1.2,2.6122622680664063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,1,power_law_1.2,3.0359423828125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,1,power_law_1.2,4.098888854980468
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,1,power_law_1.2,4.771832885742187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,1,power_law_1.2,7.139336547851562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,1,power_law_1.2,8.684942626953125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,1,power_law_1.2,0.05381631851196289
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,1,power_law_1.2,0.075447678565979
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,1,power_law_1.2,0.10779328346252441
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,1,power_law_1.2,0.17856704711914062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,1,power_law_1.2,0.2720889663696289
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,1,power_law_1.2,0.39394432067871094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,1,power_law_1.2,0.46284606933593747
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,4,1,power_law_1.2,11.21961181640625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,1,power_law_1.2,0.5949958419799805
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,1,power_law_1.2,0.6120908737182618
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,1,power_law_1.2,0.631729278564453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,1,power_law_1.2,0.6550093078613282
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,1,power_law_1.2,0.676974105834961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,1,power_law_1.2,0.6956108856201172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,1,power_law_1.2,0.7172518157958985
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,1,power_law_1.2,0.7329491424560547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,1,power_law_1.2,0.7506829071044921
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,1,power_law_1.2,0.8340505981445313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,1,power_law_1.2,0.8809158325195312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,1,power_law_1.2,0.9285529327392578
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,1,power_law_1.2,1.0160115051269532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,1,power_law_1.2,1.1137286376953124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,1,power_law_1.2,1.2767552185058595
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,1,power_law_1.2,1.4772344970703126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,1,power_law_1.2,1.7834860229492187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,1,power_law_1.2,2.2004217529296874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,1,power_law_1.2,2.8166891479492184
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,1,power_law_1.2,3.5717266845703124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,1,power_law_1.2,0.07789951801300049
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,4,1,power_law_1.2,4.153209533691406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,1,power_law_1.2,0.08211520195007324
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,1,power_law_1.2,0.08288640022277832
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,1,power_law_1.2,0.09866496086120605
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,1,power_law_1.2,0.0988595199584961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,1,power_law_1.2,0.11849599838256837
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,1,power_law_1.2,0.11796287536621093
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,1,power_law_1.2,0.1395321559906006
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,1,power_law_1.2,0.14887295722961427
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,1,power_law_1.2,0.1999020767211914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,1,power_law_1.2,0.2224166488647461
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,1,power_law_1.2,0.26954816818237304
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,1,power_law_1.2,0.3729644775390625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,1,power_law_1.2,0.0775878381729126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,1,power_law_1.2,0.07689663887023926
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,1,power_law_1.2,0.46925441741943363
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,1,power_law_1.2,0.32528446197509764
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,1,power_law_1.2,1.3021888732910156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,1,power_law_1.2,0.659939193725586
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,1,power_law_1.2,0.8659423828125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,1,power_law_1.2,2.5181549072265623
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,1,power_law_1.2,1.7007238769531248
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,1,power_law_1.2,3.3189306640625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,1,power_law_1.2,4.9557363891601565
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,1,power_law_1.01,0.04952383995056152
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,1,power_law_1.01,0.0834444808959961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,1,power_law_1.01,0.104901762008667
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,1,power_law_1.01,0.16498239517211916
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,1,power_law_1.01,0.1708358383178711
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,1,power_law_1.01,0.17598527908325196
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,1,power_law_1.01,0.1789836883544922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,1,power_law_1.01,0.18226240158081056
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,1,power_law_1.01,0.19165184020996093
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,1,power_law_1.01,0.2019468879699707
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,1,power_law_1.01,0.20573888778686525
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,1,power_law_1.01,0.21042943954467774
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,1,power_law_1.01,0.21244224548339843
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,1,power_law_1.01,0.2174662399291992
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,1,power_law_1.01,0.2514796829223633
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,1,power_law_1.01,0.2669580841064453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,1,power_law_1.01,0.2881894493103027
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,1,power_law_1.01,0.34605247497558594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,1,power_law_1.01,0.39912574768066406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,1,power_law_1.01,0.5105766296386719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,1,power_law_1.01,0.6288697433471679
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,1,power_law_1.01,0.8594163513183594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,1,power_law_1.01,1.0955955505371093
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,1,power_law_1.01,1.5997273254394533
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,1,power_law_1.01,2.063695373535156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,4,1,power_law_1.01,2.484837188720703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,1,power_law_1.2,0.03167295932769775
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,1,power_law_1.2,0.04680831909179688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,1,power_law_1.2,0.06087232112884522
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,1,power_law_1.2,0.07966591835021972
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,1,power_law_1.2,0.10002495765686034
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,1,power_law_1.2,0.16531904220581056
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,1,power_law_1.2,0.17786432266235352
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,1,power_law_1.2,0.18028608322143555
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,1,power_law_1.2,0.1841062355041504
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,1,power_law_1.2,0.1985100746154785
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,1,power_law_1.2,0.16753408432006836
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,1,power_law_1.2,0.20502336502075194
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,1,power_law_1.2,0.2144633674621582
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,1,power_law_1.2,0.21691263198852537
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,1,power_law_1.2,0.2372979164123535
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,1,power_law_1.2,0.20274816513061525
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,1,power_law_1.2,0.25503040313720704
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,1,power_law_1.2,0.26433280944824217
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,1,power_law_1.2,0.4074073410034179
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,1,power_law_1.2,0.5104755020141601
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,1,power_law_1.2,0.6343468856811524
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,1,power_law_1.2,0.3455782318115234
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,1,power_law_1.2,0.8574463653564454
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,1,power_law_1.2,1.0991302490234376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,1,power_law_1.2,1.5526995849609375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,1,power_law_1.2,2.0350445556640624
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,4,1,power_law_1.2,2.5105331420898436
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,1,power_law_1.01,0.09674240112304687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,1,power_law_1.01,0.17336128234863282
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,1,power_law_1.01,0.28653951644897463
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,1,power_law_1.01,0.4646540832519531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,1,power_law_1.01,1.018028793334961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,1,power_law_1.01,0.638021125793457
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,1,power_law_1.01,1.087256317138672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,1,power_law_1.01,1.1509439849853516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,1,power_law_1.01,1.1688307189941407
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,1,power_law_1.01,1.2343186950683593
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,1,power_law_1.01,1.2755750274658202
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,1,power_law_1.01,1.4389056396484374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,1,power_law_1.01,1.709156494140625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,1,power_law_1.01,1.9166592407226564
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,1,power_law_1.01,1.4411628723144532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,1,power_law_1.01,2.0203173828125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,1,power_law_1.01,2.4904882812499998
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,1,power_law_1.01,1.820184326171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,1,power_law_1.01,2.146300811767578
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,1,power_law_1.01,2.6785018920898436
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,1,power_law_1.01,3.4493945312500003
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,1,power_law_1.01,3.959830322265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,1,power_law_1.01,5.262451171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,1,power_law_1.01,6.8722686767578125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,1,power_law_1.01,9.901306762695313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,1,power_law_1.01,12.579813232421875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,1,power_law_1.01,17.326668701171876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,1,power_law_1.01,0.07945600032806396
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,1,power_law_1.01,0.11888511657714844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,1,power_law_1.01,0.2922515106201172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,1,power_law_1.01,0.6412723541259766
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,1,power_law_1.01,0.690750732421875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,1,power_law_1.01,0.7194745635986328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,4,1,power_law_1.01,20.013648681640625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,1,power_law_1.01,0.7473299407958984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,1,power_law_1.01,0.19003328323364257
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,1,power_law_1.01,0.7524256134033204
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,1,power_law_1.01,0.7694790649414063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,1,power_law_1.01,0.7939782714843749
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,1,power_law_1.01,0.8070028686523438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,1,power_law_1.01,0.41745983123779296
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,1,power_law_1.01,0.8493619537353515
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,1,power_law_1.01,0.8388742065429687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,1,power_law_1.01,0.8779302215576171
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,1,power_law_1.01,1.1935590362548827
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,1,power_law_1.01,1.082442855834961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,1,power_law_1.01,1.3204135131835937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,1,power_law_1.01,0.9340633392333985
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,1,power_law_1.01,1.5569894409179688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,1,power_law_1.01,1.90067138671875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,1,power_law_1.01,2.218597717285156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,1,power_law_1.01,3.799063720703125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,1,power_law_1.01,2.9537338256835937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,1,power_law_1.01,5.293890380859375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,1,power_law_1.01,6.517931518554687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,4,1,power_law_1.01,8.639281616210937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,1,power_law_1.01,0.11205504417419435
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,1,power_law_1.01,0.11288767814636232
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,1,power_law_1.01,0.1130726432800293
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,1,power_law_1.01,0.11739392280578613
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,1,power_law_1.01,0.12617664337158202
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,1,power_law_1.01,0.13432000160217286
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,1,power_law_1.01,0.14187775611877443
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,1,power_law_1.01,0.15164287567138673
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,1,power_law_1.01,0.15894720077514649
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,1,power_law_1.01,0.173940486907959
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,1,power_law_1.01,0.11387200355529785
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,1,power_law_1.01,0.21220159530639648
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,1,power_law_1.01,0.2644249534606934
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,1,power_law_1.01,0.22765247344970704
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,1,power_law_1.01,0.3181196784973145
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,1,power_law_1.01,0.44250305175781246
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,1,power_law_1.01,0.6136492919921875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,1,power_law_1.01,0.35548095703125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,1,power_law_1.01,0.7881523132324219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,1,power_law_1.01,1.1326252746582033
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,1,power_law_1.01,1.4782400512695313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,1,power_law_1.01,2.1704441833496095
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,1,power_law_1.01,2.863060607910156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,1,power_law_1.01,4.217279357910156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,1,power_law_1.01,5.593355102539062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,1,power_law_1.2,0.1877267265319824
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,1,power_law_1.2,0.28808128356933593
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,1,power_law_1.2,0.10794239997863771
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,1,power_law_1.2,0.4354143905639648
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,1,power_law_1.2,1.005194854736328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,1,power_law_1.2,0.5832896041870117
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,1,power_law_1.2,1.0823302459716797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,1,power_law_1.2,1.1245996856689453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,1,power_law_1.2,1.1789036560058594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,1,power_law_1.2,1.2285945892333985
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,1,power_law_1.2,1.30198974609375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,1,power_law_1.2,1.4175033569335938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,1,power_law_1.2,1.5005381774902342
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,1,power_law_1.2,1.7166163635253906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,1,power_law_1.2,1.8918336486816407
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,1,power_law_1.2,2.0636607360839845
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,1,power_law_1.2,2.40699462890625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,1,power_law_1.2,1.9191719055175782
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,1,power_law_1.2,2.2019282531738282
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,1,power_law_1.2,2.6295199584960938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,1,power_law_1.2,3.0677349853515627
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,1,power_law_1.2,3.868223266601562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,1,power_law_1.2,4.898486328125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,1,power_law_1.2,6.656986694335937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,1,power_law_1.2,8.817293090820312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,1,power_law_1.2,12.42010498046875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,1,power_law_1.2,0.07410560131072999
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,1,power_law_1.2,0.18715776443481444
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,1,power_law_1.2,0.11336640357971191
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,1,power_law_1.2,0.2744332885742188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,1,power_law_1.2,0.37497920989990235
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,1,power_law_1.2,0.641822738647461
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,1,power_law_1.2,0.6822994995117188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,1,power_law_1.2,0.7092691040039062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,1,power_law_1.2,0.7392486572265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,1,power_law_1.2,0.7605324554443359
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,1,power_law_1.2,0.785455322265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,1,power_law_1.2,0.8140064239501953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,1,power_law_1.2,0.8272550201416016
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,1,power_law_1.2,0.8766880035400391
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,1,power_law_1.2,15.951911621093751
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,1,power_law_1.2,0.883795166015625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,1,power_law_1.2,0.9104678344726562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,1,power_law_1.2,0.9647859191894531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,1,power_law_1.2,1.1133229064941408
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,1,power_law_1.2,1.2108799743652345
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,1,power_law_1.2,1.3736639404296875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,1,power_law_1.2,1.5617517089843749
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,1,power_law_1.2,1.9410003662109374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,1,power_law_1.2,2.2218829345703126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,1,power_law_1.2,3.0790277099609376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,1,power_law_1.2,3.8051071166992188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,1,power_law_1.2,5.003891296386719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,1,power_law_1.2,0.11271743774414063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,1,power_law_1.2,0.11271360397338867
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,1,power_law_1.2,0.11285632133483887
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,1,power_law_1.2,0.11403648376464844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,1,power_law_1.2,0.11775936126708983
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,1,power_law_1.2,6.7313995361328125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,1,power_law_1.2,0.12624575614929198
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,1,power_law_1.2,0.14285247802734374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,1,power_law_1.2,0.1510572814941406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,1,power_law_1.2,0.15920255661010743
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,4,1,power_law_1.2,19.586163330078126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,1,power_law_1.2,0.17524736404418945
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,1,power_law_1.2,0.21337215423583986
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,1,power_law_1.2,0.23025407791137695
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,1,power_law_1.2,0.26318399429321293
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,1,power_law_1.2,0.1340614414215088
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,4,1,power_law_1.2,8.349220581054688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,1,power_law_1.2,0.3181407928466797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,1,power_law_1.2,0.3538483047485351
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,1,power_law_1.2,1.133776626586914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,1,power_law_1.2,0.442017936706543
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,1,power_law_1.2,0.78714111328125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,1,power_law_1.2,0.6161286544799804
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,1,power_law_1.2,1.4779263305664063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,1,power_law_1.2,2.175237121582031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,1,power_law_1.2,2.8617709350585936
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,1,power_law_1.2,4.240796203613281
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,1,power_law_1.01,0.07230847835540771
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,1,power_law_1.01,0.06984384059906006
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,1,power_law_1.01,0.07267263889312744
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,1,power_law_1.01,0.08899711608886719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,1,power_law_1.01,0.1201740837097168
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,1,power_law_1.01,0.12087039947509766
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,1,power_law_1.01,0.1226028823852539
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,1,power_law_1.01,0.12298687934875488
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,1,power_law_1.01,0.12368448257446289
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,1,power_law_1.01,0.12515135765075686
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,1,power_law_1.01,0.12849791526794435
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,1,power_law_1.01,0.13017279624938966
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,1,power_law_1.01,0.13464703559875488
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,1,power_law_1.01,0.1413088035583496
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,1,power_law_1.01,0.14565888404846192
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,1,power_law_1.01,0.18218048095703127
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,1,power_law_1.01,0.21418687820434573
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,1,power_law_1.2,5.58604736328125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,1,power_law_1.01,0.2458745574951172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,1,power_law_1.01,0.38039230346679687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,1,power_law_1.01,0.29080448150634763
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,1,power_law_1.01,0.4682617568969727
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,1,power_law_1.01,0.6463053131103516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,1,power_law_1.01,0.8279609680175781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,1,power_law_1.01,1.1833497619628905
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,1,power_law_1.01,1.5952922058105468
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,1,power_law_1.01,2.3742834472656247
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,1,power_law_1.01,3.1890130615234376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,1,power_law_1.01,0.0443782377243042
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,1,power_law_1.01,0.04786687850952148
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,1,power_law_1.01,0.05069759845733642
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,1,power_law_1.01,0.059889922142028804
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,1,power_law_1.01,0.06283840179443359
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,1,power_law_1.01,0.06440959930419922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,1,power_law_1.01,0.06517375946044922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,1,power_law_1.01,0.06548223972320558
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,1,power_law_1.01,0.0668012809753418
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,1,power_law_1.01,0.06820096015930176
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,1,power_law_1.01,0.06860864162445068
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,1,power_law_1.01,0.058513278961181636
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,1,power_law_1.01,0.07019519805908203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,1,power_law_1.01,0.07375423908233643
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,1,power_law_1.01,0.0848755168914795
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,1,power_law_1.01,0.08939776420593262
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,1,power_law_1.01,0.09661312103271484
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,1,power_law_1.01,0.10376511573791505
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,1,power_law_1.01,0.12397503852844238
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,1,power_law_1.01,0.15211456298828124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,1,power_law_1.01,0.18398015975952148
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,1,power_law_1.01,0.24319744110107422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,1,power_law_1.01,0.3184262466430664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,1,power_law_1.01,0.4402809524536133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,8,1,power_law_1.01,3.852493591308594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,1,power_law_1.01,0.5737382507324218
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,1,power_law_1.01,0.8182329559326172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,1,power_law_1.01,1.0791212463378907
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,8,1,power_law_1.01,0.020159358978271487
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,8,1,power_law_1.01,0.03388736009597778
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,8,1,power_law_1.01,0.0355347204208374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,8,1,power_law_1.01,0.03615231990814209
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,8,1,power_law_1.01,0.03601408004760742
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,8,1,power_law_1.01,0.035763840675354
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,8,1,power_law_1.01,0.035550079345703124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,8,1,power_law_1.01,0.03647615909576416
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,8,1,power_law_1.01,0.03663680076599121
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,8,1,power_law_1.01,0.02168832063674927
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,8,1,power_law_1.01,1.3483251953125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,8,1,power_law_1.01,0.02560447931289673
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,8,1,power_law_1.01,0.0399289608001709
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,8,1,power_law_1.01,0.03619839906692505
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,8,1,power_law_1.01,0.036968960762023925
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,8,1,power_law_1.01,0.04166975975036621
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,8,1,power_law_1.01,0.04480639934539795
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,8,1,power_law_1.01,0.04647552013397217
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,8,1,power_law_1.01,0.0571014404296875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.01,0.06601984024047851
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.01,0.08598400115966796
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.01,0.10230463981628417
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.01,0.14100159645080565
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.01,0.18041536331176758
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.01,0.2641440010070801
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.01,0.34144641876220705
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,1,power_law_1.2,0.05885119915008545
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,1,power_law_1.2,0.0685920000076294
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,1,power_law_1.2,0.07014592170715332
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,1,power_law_1.2,0.08660863876342774
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,1,power_law_1.2,0.118286075592041
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,1,power_law_1.2,0.1208563232421875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,1,power_law_1.2,0.12141311645507813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,1,power_law_1.2,0.12197440147399903
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,1,power_law_1.2,0.1224505615234375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,1,power_law_1.2,0.12339648246765136
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,1,power_law_1.2,0.12632448196411133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,1,power_law_1.2,0.12859647750854492
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,1,power_law_1.2,0.13321536064147949
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,1,power_law_1.2,0.1402841567993164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,1,power_law_1.2,0.14515904426574705
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,1,power_law_1.2,0.18873151779174804
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,1,power_law_1.2,0.2114156723022461
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,1,power_law_1.2,0.24390144348144532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,1,power_law_1.2,0.2921235275268555
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.01,0.504007682800293
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,1,power_law_1.2,0.37858047485351565
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,1,power_law_1.2,0.46211967468261717
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,1,power_law_1.2,0.6402719879150391
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,1,power_law_1.2,0.814801254272461
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.01,0.6678233337402344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,1,power_law_1.2,1.1729376220703125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,8,1,power_law_1.01,0.8340799713134766
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,1,power_law_1.2,1.5556466674804688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,1,power_law_1.2,0.04093696117401123
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,1,power_law_1.2,0.0485152006149292
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,1,power_law_1.2,0.048853759765625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,1,power_law_1.2,0.06089727878570557
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,1,power_law_1.2,0.061298561096191403
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,1,power_law_1.2,0.06552127838134766
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,1,power_law_1.2,0.0644697618484497
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,1,power_law_1.2,0.06607039928436279
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,1,power_law_1.2,0.06541247844696045
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,1,power_law_1.2,0.06758592128753663
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,1,power_law_1.2,0.0667027187347412
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,1,power_law_1.2,0.07008319854736328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,1,power_law_1.2,0.06943999767303467
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,1,power_law_1.2,0.07853695869445801
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,1,power_law_1.2,0.0860108757019043
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,1,power_law_1.2,0.09923456192016603
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,1,power_law_1.2,0.0954252815246582
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,1,power_law_1.2,0.10339776039123535
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,1,power_law_1.2,0.12313407897949218
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,1,power_law_1.2,0.14847295761108398
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,1,power_law_1.2,0.18295679092407227
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,1,power_law_1.2,2.3057305908203127
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,1,power_law_1.2,0.24478271484374997
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,1,power_law_1.2,0.3170380783081055
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,1,power_law_1.2,3.0593011474609377
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,1,power_law_1.2,0.44204864501953123
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,1,power_law_1.2,0.5718726348876954
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,8,1,power_law_1.2,4.095875854492188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,1,power_law_1.2,0.8160352325439453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,8,1,power_law_1.2,0.022370560169219973
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,8,1,power_law_1.2,0.02430783987045288
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,8,1,power_law_1.2,0.03461632013320923
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,8,1,power_law_1.2,0.0341644811630249
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,8,1,power_law_1.2,0.03622976064682007
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,8,1,power_law_1.2,0.03626688003540039
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,8,1,power_law_1.2,0.03583487987518311
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,8,1,power_law_1.2,0.03658175945281982
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,8,1,power_law_1.2,0.03674240112304687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,8,1,power_law_1.2,0.03673088073730469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,8,1,power_law_1.2,0.037517440319061277
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,8,1,power_law_1.2,0.03705984115600586
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,8,1,power_law_1.2,0.03930880069732666
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,8,1,power_law_1.2,0.01975167989730835
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,8,1,power_law_1.2,0.046144638061523434
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,8,1,power_law_1.2,0.04180416107177735
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,1,power_law_1.2,1.0668198394775392
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,8,1,power_law_1.2,0.04868864059448243
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.2,0.08652223587036133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.2,0.06659711837768555
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,8,1,power_law_1.2,0.05805376052856446
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.2,0.14651455879211425
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,8,1,power_law_1.2,1.3212658691406252
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.2,0.10591103553771972
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.2,0.1860972785949707
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.2,0.27183616638183594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.2,0.3574630355834961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,1,power_law_1.01,0.07820032119750978
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,1,power_law_1.01,0.09283712387084961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,1,power_law_1.01,0.09938431739807128
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,1,power_law_1.01,0.1289139175415039
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,1,power_law_1.01,0.16497919082641602
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,1,power_law_1.01,0.16598911285400392
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,1,power_law_1.01,0.1676870346069336
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,1,power_law_1.01,0.16727039337158203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,1,power_law_1.01,0.1679840087890625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,1,power_law_1.01,0.1702195167541504
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,1,power_law_1.01,0.1811065673828125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,1,power_law_1.01,0.1917081642150879
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,1,power_law_1.01,0.20586175918579103
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,1,power_law_1.01,0.22093183517456055
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,1,power_law_1.01,0.23468223571777344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,1,power_law_1.01,0.2899488067626953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,1,power_law_1.01,0.3150553512573242
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,1,power_law_1.01,0.38444671630859373
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,1,power_law_1.01,0.4571891021728516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,1,power_law_1.01,0.6050054550170898
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,1,power_law_1.01,0.7530470275878907
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,1,power_law_1.01,1.0575174713134765
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,1,power_law_1.01,1.399310760498047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.2,0.5238131332397461
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,1,power_law_1.01,1.9884909057617186
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,1,power_law_1.01,2.848314208984375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.2,0.68661376953125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,1,power_law_1.01,3.983004150390625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,1,power_law_1.01,0.05170623779296875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,1,power_law_1.01,0.05936768054962158
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,1,power_law_1.01,0.06408383846282958
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,1,power_law_1.01,0.08351872444152833
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,1,power_law_1.01,0.08965951919555663
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,1,power_law_1.01,0.09060735702514648
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,1,power_law_1.01,0.09174079895019531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,1,power_law_1.01,0.09175423622131348
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,1,power_law_1.01,0.09234623908996582
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,1,power_law_1.01,0.09423744201660156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,1,power_law_1.01,0.09546560287475586
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,1,power_law_1.01,0.09588735580444337
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,1,power_law_1.01,0.09893247604370117
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,1,power_law_1.01,0.11278656005859375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,1,power_law_1.01,0.11819135665893554
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,1,power_law_1.01,0.11995072364807129
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,1,power_law_1.01,0.12794431686401367
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,1,power_law_1.01,0.14363840103149414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,1,power_law_1.01,0.17958784103393555
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,1,power_law_1.01,0.22821376800537108
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,1,power_law_1.01,5.69052734375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,1,power_law_1.01,0.26120576858520506
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,8,1,power_law_1.2,0.8538240051269531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,1,power_law_1.01,0.36482944488525393
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,1,power_law_1.01,0.4618182373046875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,1,power_law_1.01,0.6438495635986328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,8,1,power_law_1.01,6.801041259765626
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,1,power_law_1.01,0.8393862152099609
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,1,power_law_1.01,1.2649900817871094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,1,power_law_1.01,1.757861785888672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,8,1,power_law_1.01,2.138504943847656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,8,1,power_law_1.01,0.024333438873291015
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,8,1,power_law_1.01,0.04761343955993652
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,8,1,power_law_1.01,0.047655677795410155
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,8,1,power_law_1.01,0.04814080238342285
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,8,1,power_law_1.01,0.047700481414794924
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,8,1,power_law_1.01,0.047932162284851074
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,8,1,power_law_1.01,0.048519039154052736
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,8,1,power_law_1.01,0.049192318916320796
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,8,1,power_law_1.01,0.03092031955718994
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,8,1,power_law_1.01,0.038175361156463625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,8,1,power_law_1.01,0.04665599822998047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,8,1,power_law_1.01,0.051584000587463374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,8,1,power_law_1.01,0.050931839942932135
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,8,1,power_law_1.01,0.05646783828735351
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,8,1,power_law_1.01,0.05898047924041748
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,8,1,power_law_1.01,0.05981440067291259
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,8,1,power_law_1.01,0.08051648139953613
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,8,1,power_law_1.01,0.0666976022720337
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.01,0.09502335548400878
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.01,0.12301568031311035
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.01,0.14918911933898926
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.01,0.21147647857666016
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.01,0.268668155670166
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.01,0.3898355102539063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.01,0.5089433670043946
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,1,power_law_1.2,0.0780025577545166
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,1,power_law_1.2,0.09045568466186524
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,1,power_law_1.2,0.09467583656311035
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,1,power_law_1.2,0.1281491184234619
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,1,power_law_1.2,0.16168447494506838
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,1,power_law_1.2,0.1641043281555176
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,1,power_law_1.2,0.16487424850463867
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,1,power_law_1.2,0.1655571174621582
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,1,power_law_1.2,0.16686784744262695
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,1,power_law_1.2,0.1682431983947754
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,1,power_law_1.2,0.18027328491210937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,1,power_law_1.2,0.19154176712036133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,1,power_law_1.2,0.2031756782531738
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,1,power_law_1.2,0.2224505615234375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,1,power_law_1.2,0.24053184509277345
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,1,power_law_1.2,0.29742399215698245
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,1,power_law_1.2,0.3125209617614746
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,1,power_law_1.2,0.3829395294189453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,1,power_law_1.2,0.45836734771728516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,1,power_law_1.2,0.6071667098999024
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,1,power_law_1.2,0.7451481628417969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.01,0.7620889282226562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.01,0.9966694641113282
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,1,power_law_1.2,1.0765074920654296
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,1,power_law_1.2,1.405560302734375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,1,power_law_1.2,2.035566711425781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,8,1,power_law_1.01,1.2420787048339843
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,1,power_law_1.2,2.8182669067382813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,1,power_law_1.2,0.051756157875061035
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,1,power_law_1.2,0.058430719375610354
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,1,power_law_1.2,0.06067391872406006
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,1,power_law_1.2,0.08277055740356445
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,1,power_law_1.2,0.08753215789794921
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,1,power_law_1.2,0.0898259162902832
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,1,power_law_1.2,0.09118975639343262
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,1,power_law_1.2,0.09167872428894043
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,1,power_law_1.2,0.09256704330444336
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,1,power_law_1.2,0.09326335906982422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,1,power_law_1.2,0.09487615585327148
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,1,power_law_1.2,0.09609087944030761
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,1,power_law_1.2,0.09779135704040527
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,1,power_law_1.2,0.11162431716918945
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,1,power_law_1.2,0.12204031944274903
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,1,power_law_1.2,0.12376704216003417
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,1,power_law_1.2,0.13143296241760255
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,1,power_law_1.2,0.14376511573791503
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,1,power_law_1.2,0.17916479110717773
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,1,power_law_1.2,0.2159391975402832
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,1,power_law_1.2,0.26095167160034183
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,1,power_law_1.2,0.36132030487060546
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,1,power_law_1.2,4.288136901855468
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,1,power_law_1.2,0.45746688842773436
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,1,power_law_1.2,5.3064294433593755
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,1,power_law_1.2,0.6494067382812501
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,1,power_law_1.2,0.8280710601806641
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,8,1,power_law_1.2,6.957001953124999
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,8,1,power_law_1.2,0.024657280445098878
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,8,1,power_law_1.2,0.03004928112030029
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,1,power_law_1.2,1.2103052520751953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,8,1,power_law_1.2,0.03336832046508789
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,8,1,power_law_1.2,0.04696000099182129
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,8,1,power_law_1.2,0.046940159797668454
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,8,1,power_law_1.2,0.04757760047912597
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,8,1,power_law_1.2,0.04819647789001465
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,8,1,power_law_1.2,0.047559041976928715
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,8,1,power_law_1.2,0.04810048103332519
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,8,1,power_law_1.2,0.04840127944946289
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,8,1,power_law_1.2,0.04925183773040771
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,8,1,power_law_1.2,0.051373438835144045
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,8,1,power_law_1.2,0.057045121192932126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,8,1,power_law_1.2,0.06079232215881347
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,8,1,power_law_1.2,0.061953921318054196
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,8,1,power_law_1.2,0.050970239639282225
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,8,1,power_law_1.2,0.06710464000701905
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,1,power_law_1.2,1.7989447021484373
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,8,1,power_law_1.2,2.2107533264160155
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,8,1,power_law_1.2,0.08217599868774414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.2,0.15367232322692873
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.2,0.09547904014587402
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.2,0.12439231872558594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.2,0.4046630477905273
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.2,0.21626623153686522
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.2,0.27534400939941406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.2,0.530145263671875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,1,power_law_1.01,0.06742784023284912
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,1,power_law_1.01,0.08359359741210938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,1,power_law_1.01,0.10462207794189453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,1,power_law_1.01,0.15012096405029296
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,1,power_law_1.01,0.22350080490112303
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,1,power_law_1.01,0.30880640029907225
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,1,power_law_1.01,0.3945574569702148
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,1,power_law_1.01,0.4070758438110351
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,1,power_law_1.01,0.4147980880737305
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,1,power_law_1.01,0.42081470489501954
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,1,power_law_1.01,0.4432902526855469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.2,1.03259521484375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,1,power_law_1.01,0.4794623947143554
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,1,power_law_1.01,0.49491134643554685
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,1,power_law_1.01,0.5319289779663086
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,1,power_law_1.01,0.5658636856079101
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,1,power_law_1.01,0.6123955154418945
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,1,power_law_1.01,0.6615718078613282
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,1,power_law_1.01,0.7001612854003907
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,1,power_law_1.01,0.731259536743164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,1,power_law_1.01,0.8262111663818359
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,1,power_law_1.01,0.9527232360839843
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,1,power_law_1.01,1.1657670593261718
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,1,power_law_1.01,1.3878930664062499
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.2,0.7776019287109375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,1,power_law_1.01,1.953975067138672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,1,power_law_1.01,2.318317413330078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,1,power_law_1.01,3.347083435058594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,1,power_law_1.01,0.04806655883789063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,1,power_law_1.01,4.352435913085937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,1,power_law_1.01,0.08274239540100098
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,1,power_law_1.01,0.0547046422958374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,1,power_law_1.01,0.12356608390808106
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,1,power_law_1.01,0.191409912109375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,8,1,power_law_1.01,5.141742553710937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,1,power_law_1.01,0.24671167373657227
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,1,power_law_1.01,0.31858751296997073
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,1,power_law_1.01,0.3273484802246094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,1,power_law_1.01,0.3341772842407226
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,1,power_law_1.01,0.3385561752319336
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,1,power_law_1.01,0.3473158264160156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,1,power_law_1.01,0.3607526397705078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,1,power_law_1.01,0.3663475036621094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,1,power_law_1.01,0.37117504119873046
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,1,power_law_1.01,0.4216704177856445
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,1,power_law_1.01,0.42540927886962887
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,1,power_law_1.01,0.4450054550170899
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,1,power_law_1.01,0.47865345001220705
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,1,power_law_1.01,0.5063942337036134
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,1,power_law_1.01,0.5554431915283203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,1,power_law_1.01,0.6274303817749024
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,1,power_law_1.01,0.7429798126220704
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,1,power_law_1.01,0.8705868530273438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,1,power_law_1.01,1.1396076965332032
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,1,power_law_1.01,1.418914489746094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,1,power_law_1.01,1.9262342834472654
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,1,power_law_1.01,2.4809068298339847
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,8,1,power_law_1.01,2.9909912109375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,1,power_law_1.01,0.06284863948822021
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,1,power_law_1.01,0.10267775535583497
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,1,power_law_1.01,0.0825004768371582
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,1,power_law_1.01,0.1280799961090088
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,8,1,power_law_1.2,1.2810354614257813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,1,power_law_1.01,0.15246912002563479
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,1,power_law_1.01,0.19215551376342774
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,1,power_law_1.01,0.06815872192382813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,1,power_law_1.01,0.2070790481567383
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,1,power_law_1.01,0.2108633613586426
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,1,power_law_1.01,0.22017343521118163
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,1,power_law_1.01,0.23252864837646486
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,1,power_law_1.01,0.22826303482055663
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,1,power_law_1.01,0.24256895065307615
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,1,power_law_1.01,0.254531192779541
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,1,power_law_1.01,0.2631167984008789
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,1,power_law_1.01,0.2845305633544922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,1,power_law_1.01,0.2012607955932617
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,1,power_law_1.01,0.330458869934082
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.01,0.37947456359863285
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.01,0.4686240005493164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.01,0.5805465698242187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.01,0.7620979309082031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.01,0.9848108673095703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.01,1.3927290344238281
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.01,1.900767364501953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,1,power_law_1.2,0.06668416023254395
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,1,power_law_1.2,0.0846886444091797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,1,power_law_1.2,0.10553855895996093
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,1,power_law_1.2,0.1450476837158203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,1,power_law_1.2,0.2116179275512695
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,8,1,power_law_1.01,3.763059692382812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,8,1,power_law_1.01,2.832796936035156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,1,power_law_1.2,0.3090681648254395
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,1,power_law_1.2,0.3875968170166016
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,1,power_law_1.2,0.4002707290649414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,1,power_law_1.2,0.40859840393066404
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,1,power_law_1.2,0.4224070358276367
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,1,power_law_1.2,0.4420691299438476
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,1,power_law_1.2,0.4792851257324219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,1,power_law_1.2,0.5029580688476563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,1,power_law_1.2,0.5400441741943359
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,1,power_law_1.2,0.6112755203247071
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,1,power_law_1.2,0.6362464141845703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,1,power_law_1.2,0.6677734375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,1,power_law_1.2,0.7042931365966797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,1,power_law_1.2,0.7411539459228516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,1,power_law_1.2,0.8619058990478516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,1,power_law_1.2,0.9628243255615235
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,1,power_law_1.2,1.2012947082519532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,1,power_law_1.2,1.4221420288085938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,1,power_law_1.2,1.8709452819824217
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,1,power_law_1.2,2.412097930908203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,1,power_law_1.2,3.3199755859375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,1,power_law_1.2,0.04771520137786865
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,1,power_law_1.2,0.058166399002075195
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,1,power_law_1.2,0.08346112251281738
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,1,power_law_1.2,0.11926336288452148
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,1,power_law_1.2,4.21493896484375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,1,power_law_1.2,0.18397504806518555
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,1,power_law_1.2,0.31527296066284183
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,1,power_law_1.2,0.32410430908203125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,8,1,power_law_1.2,5.36022216796875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,1,power_law_1.2,0.33126785278320314
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,1,power_law_1.2,0.33698368072509766
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,1,power_law_1.2,0.3467071914672851
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,7168,2048,8,256,8,1,power_law_1.01,4.586046142578125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,1,power_law_1.2,0.3562464141845703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,1,power_law_1.2,0.3666739273071289
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,1,power_law_1.2,0.3812249755859375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,1,power_law_1.2,0.4239641571044922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,1,power_law_1.2,0.44069438934326166
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,1,power_law_1.2,0.4583942413330078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,1,power_law_1.2,0.4905100631713867
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,1,power_law_1.2,0.5208492660522461
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,1,power_law_1.2,0.5823788833618164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,1,power_law_1.2,0.6434719848632813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,1,power_law_1.2,0.7864710235595703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,1,power_law_1.2,0.9183103942871094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,1,power_law_1.2,0.24047679901123048
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,1,power_law_1.2,1.1491315460205078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,1,power_law_1.2,1.4374879455566405
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,1,power_law_1.2,1.957884826660156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,1,power_law_1.2,0.06310080051422119
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,1,power_law_1.2,2.4742681884765623
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,1,power_law_1.2,0.08016063690185546
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,1,power_law_1.2,0.0689024019241333
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,1,power_law_1.2,0.09918144226074219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,8,1,power_law_1.2,2.9897189331054688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,1,power_law_1.2,0.12204992294311523
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,1,power_law_1.2,0.14788031578063965
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,1,power_law_1.2,0.19047679901123046
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,1,power_law_1.2,0.19826879501342773
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,1,power_law_1.2,0.20629888534545898
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,1,power_law_1.2,0.21268032073974608
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,1,power_law_1.2,0.21991872787475586
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,1,power_law_1.2,0.22840448379516604
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,1,power_law_1.2,0.23755071640014647
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,1,power_law_1.2,0.2542156791687012
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,1,power_law_1.2,0.2685600090026855
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,1,power_law_1.2,0.2817856025695801
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,1,power_law_1.2,0.31456512451171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.2,0.4493331146240235
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.2,0.5946969604492187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.2,0.7419961547851562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,1,power_law_1.2,0.3787788772583008
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.2,1.1168978881835936
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.2,1.485590362548828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.2,2.180614471435547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.2,3.0039962768554687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,1,power_law_1.01,0.08626303672790528
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,1,power_law_1.01,0.09082240104675293
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,1,power_law_1.01,0.12059904098510743
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,1,power_law_1.01,0.1259385585784912
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,1,power_law_1.01,0.12987327575683594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,1,power_law_1.01,0.13633472442626954
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,1,power_law_1.01,0.1400972843170166
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,1,power_law_1.01,0.1515897560119629
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,1,power_law_1.01,0.16566528320312499
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,1,power_law_1.01,0.1687788772583008
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,1,power_law_1.01,0.17414848327636717
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,1,power_law_1.01,0.17841535568237304
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,1,power_law_1.01,0.18205888748168947
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,1,power_law_1.01,0.1926131248474121
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,1,power_law_1.01,0.22565887451171873
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,1,power_law_1.01,0.27927360534667967
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,1,power_law_1.01,0.32291072845458985
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,1,power_law_1.01,0.3844160079956055
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,1,power_law_1.01,0.5037497711181641
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,1,power_law_1.01,0.624911994934082
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,1,power_law_1.01,0.8651846313476563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,1,power_law_1.01,1.1030496215820311
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,1,power_law_1.01,1.5805996704101564
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,1,power_law_1.01,2.0567718505859376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,8,1,power_law_1.01,2.530082550048828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,1,power_law_1.01,0.037179520130157476
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,1,power_law_1.01,0.04460608005523682
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,1,power_law_1.01,0.05749695777893067
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,1,power_law_1.01,0.07036928176879884
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,1,power_law_1.01,0.08972607612609863
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,1,power_law_1.01,0.11876864433288574
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,1,power_law_1.01,0.12599360466003418
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,1,power_law_1.01,0.12849791526794435
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,1,power_law_1.01,0.13195903778076173
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,1,power_law_1.01,0.13293375968933105
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,1,power_law_1.01,0.13581695556640624
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,1,power_law_1.01,0.15356672286987305
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,1,power_law_1.01,0.15595840454101562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,1,power_law_1.01,0.15970751762390137
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,1,power_law_1.01,0.16285568237304687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,1,power_law_1.01,0.1644326400756836
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,1,power_law_1.01,0.1748262405395508
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,1,power_law_1.01,0.18778303146362305
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,1,power_law_1.01,0.2150943946838379
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,1,power_law_1.01,0.24883199691772462
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,1,power_law_1.01,0.2916671943664551
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,1,power_law_1.01,0.36962432861328126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,1,power_law_1.01,0.4484902572631836
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,1,power_law_1.01,0.6165273666381836
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,1,power_law_1.01,0.7702336120605469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,1,power_law_1.01,1.0995225524902343
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,1,power_law_1.01,1.41745849609375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,8,1,power_law_1.01,1.7387557983398438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,1,power_law_1.01,0.04270912170410156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,1,power_law_1.01,0.06624767780303956
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,1,power_law_1.01,0.08236288070678711
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,1,power_law_1.01,0.08613696098327636
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,1,power_law_1.01,0.08782784461975097
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,1,power_law_1.01,0.08990400314331055
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,1,power_law_1.01,0.09118911743164063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,1,power_law_1.01,0.09485055923461913
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,1,power_law_1.01,0.0971225643157959
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,1,power_law_1.01,0.09912768363952637
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,1,power_law_1.01,0.10554688453674317
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,1,power_law_1.01,0.04937600135803223
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,1,power_law_1.01,0.04477248191833496
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,1,power_law_1.01,0.056782717704772945
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,1,power_law_1.01,0.11443903923034668
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,1,power_law_1.01,0.1280083179473877
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,1,power_law_1.01,0.15215999603271485
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,1,power_law_1.01,0.11196415901184083
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,1,power_law_1.01,0.18183359146118164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,1,power_law_1.01,0.2987334442138672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,1,power_law_1.01,0.2409286308288574
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,1,power_law_1.01,0.4300409698486328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,1,power_law_1.01,0.5556070327758789
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,1,power_law_1.01,0.8120460510253906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,1,power_law_1.01,1.0986803436279298
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,1,power_law_1.2,0.03659071922302246
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,1,power_law_1.01,1.6558015441894534
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,1,power_law_1.2,0.06466944217681884
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,1,power_law_1.2,0.07372032165527344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,1,power_law_1.2,0.08674816131591798
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,1,power_law_1.2,0.12210623741149902
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,1,power_law_1.2,0.12798080444335938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,1,power_law_1.2,0.05384831905364991
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,1,power_law_1.2,0.13405311584472657
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,1,power_law_1.2,0.13865471839904786
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,1,power_law_1.2,0.14228544235229493
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,1,power_law_1.2,0.15133440017700195
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,1,power_law_1.2,0.16701631546020507
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,1,power_law_1.2,0.1678348731994629
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,1,power_law_1.2,0.17640832901000975
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,1,power_law_1.2,0.1783660888671875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,1,power_law_1.2,0.18385280609130858
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,1,power_law_1.2,0.19277311325073243
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,1,power_law_1.2,0.2277337646484375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,1,power_law_1.2,0.2770988845825195
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,1,power_law_1.2,0.3246489715576172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,1,power_law_1.2,0.3800435256958008
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,1,power_law_1.2,0.5054195022583008
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,1,power_law_1.01,2.2645318603515627
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,1,power_law_1.2,0.621443214416504
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,8,1,power_law_1.01,2.847850341796875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,1,power_law_1.2,0.8680089569091797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,1,power_law_1.2,1.1070252990722655
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,1,power_law_1.2,0.03670527935028076
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,1,power_law_1.2,0.04746880054473877
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,1,power_law_1.2,0.05662271976470947
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,1,power_law_1.2,0.0672870397567749
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,1,power_law_1.2,0.08675071716308594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,1,power_law_1.2,0.12075648307800294
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,1,power_law_1.2,0.12678848266601564
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,1,power_law_1.2,0.12987263679504396
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,1,power_law_1.2,0.1319385623931885
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,1,power_law_1.2,0.13296768188476563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,1,power_law_1.2,1.5718995666503905
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,1,power_law_1.2,0.1348806381225586
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,1,power_law_1.2,0.15405823707580565
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,1,power_law_1.2,0.15663488388061525
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,1,power_law_1.2,0.16169727325439454
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,1,power_law_1.2,0.16505855560302735
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,1,power_law_1.2,0.17253568649291992
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,1,power_law_1.2,0.1759699249267578
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,1,power_law_1.2,0.1927724838256836
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,1,power_law_1.2,0.22064319610595703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,1,power_law_1.2,0.2563878440856934
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,1,power_law_1.2,0.2972006416320801
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,1,power_law_1.2,0.3696780776977539
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,1,power_law_1.2,0.4543366241455078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,1,power_law_1.2,0.6145267105102539
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,1,power_law_1.2,0.7736966705322266
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,1,power_law_1.2,2.0467910766601562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,1,power_law_1.2,1.0942649841308594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,1,power_law_1.2,0.04267136096954346
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,1,power_law_1.2,0.04642687797546387
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,1,power_law_1.2,0.0501529598236084
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,1,power_law_1.2,0.05571263790130615
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,1,power_law_1.2,0.06407999992370605
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,1,power_law_1.2,0.0828825569152832
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,1,power_law_1.2,0.0867750358581543
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,1,power_law_1.2,0.08896703720092773
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,1,power_law_1.2,0.09099136352539063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,1,power_law_1.2,0.09264639854431153
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,1,power_law_1.2,0.09591872215270995
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,1,power_law_1.2,0.09965312004089355
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,1,power_law_1.2,0.10288000106811523
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,1,power_law_1.2,0.11078656196594239
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,1,power_law_1.2,0.11876992225646972
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,1,power_law_1.2,0.12690624237060547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,1,power_law_1.2,1.4216294860839844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,1,power_law_1.2,0.14469311714172364
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,1,power_law_1.2,0.18219520568847655
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,8,1,power_law_1.2,2.5387359619140626
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,1,power_law_1.2,0.22421951293945314
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,1,power_law_1.2,0.31104576110839843
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,1,power_law_1.2,0.40461376190185544
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,1,power_law_1.2,0.5893868637084961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,8,1,power_law_1.2,1.7323207092285156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,1,power_law_1.2,0.790545883178711
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,1,power_law_1.2,1.175839385986328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,1,power_law_1.2,1.5531513977050782
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,1,power_law_1.01,0.06891647815704346
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,1,power_law_1.01,0.08347711563110352
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,1,power_law_1.01,0.10428480148315429
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,1,power_law_1.01,0.14243264198303224
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,1,power_law_1.01,0.1895942306518555
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,1,power_law_1.01,0.302744312286377
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,1,power_law_1.01,0.31487871170043946
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,1,power_law_1.01,0.3243628692626953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,1,power_law_1.01,0.3315398406982422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,1,power_law_1.01,0.33342784881591797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,1,power_law_1.01,0.3464396667480469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,1,power_law_1.01,0.3737190246582031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,1,power_law_1.01,0.40192832946777346
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,1,power_law_1.01,0.4147257614135742
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,1,power_law_1.01,0.4211936187744141
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,1,power_law_1.01,0.43578880310058593
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,1,power_law_1.01,0.5022272109985352
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,1,power_law_1.01,0.547817611694336
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,1,power_law_1.01,0.5813868713378907
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,1,power_law_1.01,0.675827178955078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,1,power_law_1.01,0.8327718353271484
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,1,power_law_1.01,1.0311622619628906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,1,power_law_1.2,2.321677398681641
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,1,power_law_1.01,1.2604422760009766
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,1,power_law_1.01,1.8133517456054686
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,1,power_law_1.01,2.22359619140625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,1,power_law_1.01,3.434627075195313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,1,power_law_1.01,4.159518737792969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,1,power_law_1.01,0.04762303829193115
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,1,power_law_1.01,0.06160128116607666
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,1,power_law_1.01,0.08380543708801269
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,1,power_law_1.01,0.11069760322570801
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,1,power_law_1.01,0.15995264053344727
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,1,power_law_1.01,0.22448192596435548
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,1,power_law_1.01,0.23532800674438475
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,1,power_law_1.01,0.24110784530639648
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,8,1,power_law_1.01,5.33632080078125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,1,power_law_1.01,0.24579776763916014
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,1,power_law_1.01,0.24919424057006836
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,1,power_law_1.01,0.25414400100708007
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,1,power_law_1.01,0.25987136840820313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,1,power_law_1.01,0.3002758407592773
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,1,power_law_1.01,0.31591039657592773
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,1,power_law_1.01,0.31755456924438474
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,1,power_law_1.01,0.30695615768432616
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,1,power_law_1.01,0.3404185485839844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,1,power_law_1.01,0.3656889724731445
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,1,power_law_1.01,0.3888313674926758
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,1,power_law_1.01,0.4553881454467773
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,1,power_law_1.01,0.5187392044067383
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,1,power_law_1.01,0.657606430053711
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,1,power_law_1.01,0.7908198547363281
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,1,power_law_1.01,1.058402557373047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,1,power_law_1.01,1.3314598083496094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,1,power_law_1.01,1.855623016357422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,1,power_law_1.01,2.393296661376953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,8,1,power_law_1.01,2.9166445922851563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,1,power_law_1.2,3.095287170410156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,1,power_law_1.01,0.047249279022216796
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,1,power_law_1.01,0.0813651180267334
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,1,power_law_1.01,0.05379263877868652
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,1,power_law_1.01,0.09839872360229493
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,1,power_law_1.01,0.13710271835327148
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,1,power_law_1.01,0.14390463829040528
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,1,power_law_1.01,0.1482271957397461
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,1,power_law_1.01,0.15215807914733886
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,1,power_law_1.01,0.15521408081054688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,1,power_law_1.01,0.16145792007446289
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,1,power_law_1.01,0.16701183319091797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,1,power_law_1.01,0.17133375167846682
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,1,power_law_1.01,0.18198144912719727
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,1,power_law_1.01,0.19028608322143553
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,1,power_law_1.01,0.19304191589355468
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,1,power_law_1.01,0.06830143928527832
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,1,power_law_1.01,0.2571392059326172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,1,power_law_1.01,0.21791103363037106
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,1,power_law_1.01,0.2985420799255371
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,8,1,power_law_1.2,3.8733010864257813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,1,power_law_1.01,0.3799545669555664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,1,power_law_1.01,0.6605900573730469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,1,power_law_1.01,0.4741830444335937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,1,power_law_1.01,1.2476223754882814
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,1,power_law_1.01,2.4992486572265626
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,1,power_law_1.01,0.8582252502441406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,1,power_law_1.01,1.6636639404296876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,1,power_law_1.2,0.06639616012573242
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,1,power_law_1.2,0.1019916820526123
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,1,power_law_1.2,0.13303423881530763
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,1,power_law_1.2,0.08362112045288086
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,1,power_law_1.2,0.1791756820678711
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,1,power_law_1.2,0.31084096908569336
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,1,power_law_1.2,0.3133132743835449
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,1,power_law_1.2,0.32638465881347656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,1,power_law_1.2,0.3406304168701172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,1,power_law_1.2,0.335494384765625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,1,power_law_1.2,0.35803134918212887
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,1,power_law_1.2,0.3921753692626953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,1,power_law_1.2,0.40727809906005863
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,1,power_law_1.2,0.4187795257568359
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,1,power_law_1.2,0.4358451080322266
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,1,power_law_1.2,0.4545369720458984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,1,power_law_1.2,0.5138924789428712
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,1,power_law_1.2,0.555736312866211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,1,power_law_1.2,0.5888166427612305
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,1,power_law_1.2,0.6989024353027344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,1,power_law_1.2,0.8126854705810548
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,1,power_law_1.2,1.057738265991211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,1,power_law_1.2,1.2841433715820312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,1,power_law_1.2,1.796727752685547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,1,power_law_1.2,2.2507557678222656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,1,power_law_1.2,3.3279483032226564
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,1,power_law_1.01,3.3598541259765624
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,1,power_law_1.2,0.04639935970306396
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,1,power_law_1.2,0.06145408153533936
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,1,power_law_1.2,0.0816096019744873
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,1,power_law_1.2,0.10411904335021974
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,1,power_law_1.2,0.15371711730957033
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,1,power_law_1.2,0.21986303329467774
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,2560,8,160,8,1,power_law_1.01,4.233763732910156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,1,power_law_1.2,0.2331532859802246
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,1,power_law_1.2,0.2430054473876953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,1,power_law_1.2,0.24702400207519531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,1,power_law_1.2,0.2526348876953125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,1,power_law_1.2,0.2577644729614258
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,1,power_law_1.2,0.26317184448242187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,1,power_law_1.2,0.31796287536621093
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,1,power_law_1.2,0.3188262367248535
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,1,power_law_1.2,0.33075134277343754
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,1,power_law_1.2,0.3384819030761719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,1,power_law_1.2,0.35224449157714843
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,1,power_law_1.2,0.37049407958984376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,1,power_law_1.2,0.3968659210205078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,1,power_law_1.2,0.46838912963867185
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,1,power_law_1.2,0.5357363128662109
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,1,power_law_1.2,0.6755296325683593
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,1,power_law_1.2,4.156941528320313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,1,power_law_1.2,0.7993555450439453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,1,power_law_1.2,1.068383331298828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,1,power_law_1.2,1.347328643798828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,1,power_law_1.2,1.8722682189941406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,1,power_law_1.2,0.047449598312377925
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,1,power_law_1.2,0.053951358795166014
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,1,power_law_1.2,0.06574336051940918
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,1,power_law_1.2,0.07762495994567871
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,1,power_law_1.2,0.09525952339172364
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,1,power_law_1.2,0.13515904426574707
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,1,power_law_1.2,0.14399744033813477
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,1,power_law_1.2,0.1503878402709961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,1,power_law_1.2,0.1559488010406494
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,1,power_law_1.2,0.15909631729125978
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,1,power_law_1.2,2.3956147766113283
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,1,power_law_1.2,0.16622976303100584
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,1,power_law_1.2,0.17097984313964845
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,1,power_law_1.2,0.17960447311401367
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,1,power_law_1.2,0.19033599853515626
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,1,power_law_1.2,0.21911104202270507
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,8,1,power_law_1.2,5.356751098632812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,8,1,power_law_1.2,2.962215576171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,1,power_law_1.2,0.2055948829650879
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,1,power_law_1.2,0.24548479080200197
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,1,power_law_1.2,0.3019513511657715
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,1,power_law_1.2,0.3658419036865234
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,1,power_law_1.2,0.49829185485839844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,1,power_law_1.2,0.6443660736083985
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,1,power_law_1.2,0.9627500915527344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,1,power_law_1.2,1.3042253112792968
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,1,power_law_1.2,2.0053503417968748
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,1,power_law_1.2,2.7356484985351566
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,1,power_law_1.01,0.06936960220336914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,1,power_law_1.01,0.08885120391845704
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,1,power_law_1.01,0.11057984352111816
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,1,power_law_1.01,0.1570841598510742
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,1,power_law_1.01,0.2890847969055176
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,1,power_law_1.01,0.38703807830810544
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,1,power_law_1.01,0.4214630508422852
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,1,power_law_1.01,0.5317337417602539
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,1,power_law_1.01,0.5520486450195312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,1,power_law_1.01,0.56923583984375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,1,power_law_1.01,0.5846732711791992
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,1,power_law_1.01,0.5917587280273438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,1,power_law_1.01,0.6193574523925781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,1,power_law_1.01,0.6681874847412109
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,1,power_law_1.01,0.7022598266601563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,1,power_law_1.01,0.7495961761474609
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,1,power_law_1.01,0.8131635284423828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,1,power_law_1.01,0.8498617553710938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,1,power_law_1.01,0.9597164916992188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,1,power_law_1.01,1.0423935699462892
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,1,power_law_1.01,1.1270803070068358
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,1,power_law_1.2,4.2141748046875005
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,1,power_law_1.01,1.411212158203125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,1,power_law_1.01,1.6452429199218748
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,1,power_law_1.01,2.1946144104003906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,1,power_law_1.01,2.694159851074219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,1,power_law_1.01,3.654839172363281
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,1,power_law_1.01,0.050517759323120116
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,1,power_law_1.2,5.667011108398437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,1,power_law_1.01,0.057645440101623535
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,1,power_law_1.01,0.13089728355407715
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,1,power_law_1.01,0.09976703643798827
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,1,power_law_1.01,0.2147257614135742
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,1,power_law_1.01,0.3183404731750488
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,1,power_law_1.01,0.3582815933227539
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,1,power_law_1.01,0.4326355361938477
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,1,power_law_1.01,0.4450867080688477
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,8,1,power_law_1.01,5.721785888671875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,1,power_law_1.01,0.45870014190673825
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,1,power_law_1.01,0.47024063110351566
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,1,power_law_1.01,0.4788051223754882
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,1,power_law_1.01,0.49486015319824217
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,1,power_law_1.01,0.5120326232910156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,1,power_law_1.01,0.5188985443115234
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,1,power_law_1.01,0.5376287841796875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,1,power_law_1.01,0.5994220733642578
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,1,power_law_1.01,0.628210563659668
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,1,power_law_1.01,0.6553465270996094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,1,power_law_1.01,0.7136153411865235
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,1,power_law_1.01,0.7951155090332032
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,1,power_law_1.01,0.9181996917724609
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,1,power_law_1.01,1.0463967895507813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,1,power_law_1.01,1.3379916381835937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,1,power_law_1.01,1.6038124084472656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,1,power_law_1.01,2.1439750671386717
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,1,power_law_1.01,2.7034124755859374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,8,1,power_law_1.01,3.2522412109374996
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,1,power_law_1.01,0.07904640197753907
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,1,power_law_1.01,0.07958528041839599
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,1,power_law_1.01,0.08112447738647462
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,1,power_law_1.01,0.08455743789672851
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,1,power_law_1.01,0.09111616134643555
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,1,power_law_1.01,0.07876416206359863
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,1,power_law_1.01,0.09831232070922852
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,1,power_law_1.01,0.10488063812255859
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,1,power_law_1.01,0.11805631637573241
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,1,power_law_1.01,0.1142470359802246
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,1,power_law_1.01,0.13261055946350098
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,1,power_law_1.01,0.181658878326416
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,1,power_law_1.01,0.2476255989074707
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,1,power_law_1.01,0.208590087890625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,1,power_law_1.01,0.3010713577270508
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,1,power_law_1.01,0.34410240173339846
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,1,power_law_1.01,0.4346035385131836
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,1,power_law_1.01,0.6159199905395508
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,1,power_law_1.01,0.7963558197021484
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,1,power_law_1.01,1.163787536621094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,1,power_law_1.01,2.3052581787109374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,1,power_law_1.01,1.5506163024902344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,1,power_law_1.01,3.038371276855469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,1,power_law_1.2,0.07344895839691162
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,1,power_law_1.2,0.08774080276489257
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,1,power_law_1.2,0.10983424186706543
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,1,power_law_1.2,0.15622207641601563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,1,power_law_1.2,0.24168319702148439
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,1,power_law_1.2,0.3644915390014648
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,1,power_law_1.2,0.4011910247802734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,1,power_law_1.2,0.5236012649536133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,1,power_law_1.2,0.539381103515625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,1,power_law_1.01,4.537496948242188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,1,power_law_1.2,0.5544652938842773
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,1,power_law_1.2,0.5715161514282227
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,1,power_law_1.2,0.5974188613891601
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,1,power_law_1.2,0.6201011276245116
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,1,power_law_1.2,0.674361572265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,1,power_law_1.2,0.7118950653076171
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,1,power_law_1.2,0.7599929809570313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,1,power_law_1.2,0.8120467376708984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,1,power_law_1.2,0.8857676696777343
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,1,power_law_1.2,0.9655238342285155
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,1,power_law_1.2,1.0543641662597656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,1,power_law_1.2,1.1710431671142578
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,1,power_law_1.2,1.4153395080566407
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,1,power_law_1.2,1.7599519348144532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,1,power_law_1.2,2.164869079589844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,1,power_law_1.2,2.67235595703125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,1,power_law_1.2,3.5658547973632815
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,1,power_law_1.2,4.627010498046875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,1,power_law_1.2,0.05090816020965576
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,1,power_law_1.2,0.05742400169372559
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,1,power_law_1.2,0.08825535774230957
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,1,power_law_1.2,0.1310201644897461
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,1,power_law_1.2,0.20470464706420896
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,8,1,power_law_1.2,5.515106201171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,1,power_law_1.2,0.2946854400634765
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,1,power_law_1.2,0.3460230255126953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,1,power_law_1.2,0.4268544006347656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,1,power_law_1.2,0.43989631652832034
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,1,power_law_1.2,0.44765888214111327
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,1,power_law_1.2,0.46160064697265624
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,1,power_law_1.2,0.47517631530761717
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,1,power_law_1.2,0.4899155044555664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,1,power_law_1.2,0.5071705627441406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,1,power_law_1.2,0.5191001510620117
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,1,power_law_1.2,0.533427848815918
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,1,power_law_1.2,0.6021081542968749
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,1,power_law_1.2,0.6357030487060547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,1,power_law_1.2,0.6673171234130859
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,1,power_law_1.2,0.7441657257080079
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,1,power_law_1.2,0.8197119903564453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,1,power_law_1.2,0.9576518249511718
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,1,power_law_1.2,1.1041862487792968
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,1,power_law_1.2,1.3639698791503907
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,1,power_law_1.2,1.6470553588867187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,1,power_law_1.2,2.182745666503906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,1,power_law_1.2,2.75486083984375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,1,power_law_1.2,0.07879168033599854
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,8,1,power_law_1.2,3.2709100341796877
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,1,power_law_1.2,0.07886335849761963
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,1,power_law_1.2,0.07971327781677247
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,1,power_law_1.2,0.08134016036987304
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,1,power_law_1.2,0.09106111526489258
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,1,power_law_1.2,0.09956607818603516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,1,power_law_1.2,0.10505151748657227
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,1,power_law_1.2,0.11301312446594239
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,1,power_law_1.2,0.11833919525146483
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,1,power_law_1.2,0.1852889633178711
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,1,power_law_1.2,0.13503616333007812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,1,power_law_1.2,0.20602880477905275
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,1,power_law_1.2,0.24572799682617186
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,1,power_law_1.2,0.30267648696899413
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,1,power_law_1.2,0.08336640357971191
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,1,power_law_1.2,0.4332825469970703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,1,power_law_1.2,0.3427116775512695
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,1,power_law_1.2,0.6183769607543945
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,1,power_law_1.2,0.796789779663086
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,1,power_law_1.2,1.1663731384277345
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,1,power_law_1.2,1.5382284545898437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,1,power_law_1.2,2.31044677734375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,1,power_law_1.2,3.0410253906249998
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,1,power_law_1.01,0.032458240985870364
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,1,power_law_1.01,0.03821055889129639
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,1,power_law_1.01,0.06356224060058593
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,1,power_law_1.2,4.521299133300781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,1,power_law_1.01,0.08616448402404785
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,1,power_law_1.01,0.11875200271606445
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,1,power_law_1.01,0.15565823554992675
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,1,power_law_1.01,0.17034624099731446
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,1,power_law_1.01,0.1757516860961914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,1,power_law_1.01,0.18306688308715818
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,1,power_law_1.01,0.16492927551269532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,1,power_law_1.01,0.20341375350952146
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,1,power_law_1.01,0.21930559158325194
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,1,power_law_1.01,0.22160575866699217
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,1,power_law_1.01,0.2293382453918457
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,1,power_law_1.01,0.2329292869567871
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,1,power_law_1.01,0.24289600372314454
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,1,power_law_1.01,0.26307327270507813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,1,power_law_1.01,0.30931520462036133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,1,power_law_1.01,0.3396031951904297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,1,power_law_1.01,0.40009601593017574
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,1,power_law_1.01,0.4711929702758789
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,1,power_law_1.01,0.6031391906738282
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,1,power_law_1.01,0.7382169342041015
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,1,power_law_1.01,1.013614044189453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,1,power_law_1.01,1.2789842987060547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,1,power_law_1.01,1.814466552734375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,1,power_law_1.01,2.36724609375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,8,1,power_law_1.01,2.8821401977539063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,1,power_law_1.2,0.048610558509826665
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,1,power_law_1.2,0.06334335803985595
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,1,power_law_1.2,0.08054656028747559
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,1,power_law_1.2,0.03375744104385376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,1,power_law_1.2,0.10126399993896484
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,1,power_law_1.2,0.1596614360809326
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,1,power_law_1.2,0.16662656784057617
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,1,power_law_1.2,0.17475135803222658
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,1,power_law_1.2,0.18056447982788087
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,1,power_law_1.2,0.1913817596435547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,1,power_law_1.2,0.20762559890747073
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,1,power_law_1.2,0.2239481544494629
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,1,power_law_1.2,0.22802495956420898
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,1,power_law_1.2,0.23756799697875977
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,1,power_law_1.2,0.24271039962768554
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,1,power_law_1.2,0.2532352066040039
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,1,power_law_1.2,0.2806752014160156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,1,power_law_1.2,0.31629568099975586
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,1,power_law_1.2,0.34333057403564454
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,1,power_law_1.2,0.40873535156249996
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,1,power_law_1.2,0.4776704025268555
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,1,power_law_1.2,0.6094892883300782
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,1,power_law_1.2,0.7398592376708985
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,1,power_law_1.2,1.0135616302490233
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,1,power_law_1.2,1.2801933288574219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,1,power_law_1.2,1.8303974914550782
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,1,power_law_1.2,2.3564076232910156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,1,power_law_1.01,0.10458175659179687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,1,power_law_1.01,0.18040639877319337
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,1,power_law_1.01,0.07530943870544433
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,1,power_law_1.01,0.2677984046936035
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,1,power_law_1.01,0.5831955337524415
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,1,power_law_1.01,0.37737918853759767
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,1,power_law_1.01,0.6206195068359375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,1,power_law_1.01,0.6407955169677735
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,1,power_law_1.01,0.6637689971923828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,1,power_law_1.01,0.6938944244384766
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,1,power_law_1.01,0.7250713348388672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,1,power_law_1.01,0.7875667572021484
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,1,power_law_1.01,0.8273792266845703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,1,power_law_1.01,0.9543801879882812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,1,power_law_1.01,1.0069766235351563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,1,power_law_1.01,1.1323564910888673
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,1,power_law_1.01,1.3204556274414063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,8,1,power_law_1.2,2.9125375366210937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,1,power_law_1.01,1.0534233856201172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,1,power_law_1.01,1.113395233154297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,1,power_law_1.01,2.225006103515625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,1,power_law_1.01,1.657392578125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,1,power_law_1.01,1.4292608642578126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,1,power_law_1.01,2.618826904296875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,1,power_law_1.01,4.078169555664063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,1,power_law_1.01,4.69531982421875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,1,power_law_1.01,0.059852161407470704
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,1,power_law_1.01,7.324418334960937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,1,power_law_1.01,0.0849510383605957
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,1,power_law_1.01,0.20869312286376954
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,1,power_law_1.01,0.28965696334838864
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,1,power_law_1.01,0.12913215637207032
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,1,power_law_1.01,9.076068115234374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,1,power_law_1.01,0.44154304504394537
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,1,power_law_1.01,0.4676409530639648
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,1,power_law_1.01,0.4843231964111328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,1,power_law_1.01,0.5071187210083008
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,1,power_law_1.01,0.5105855941772461
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,1,power_law_1.01,0.5205126571655273
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,1,power_law_1.01,0.5460601425170898
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,1,power_law_1.01,0.5501420974731446
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,1,power_law_1.01,0.5763040161132812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,1,power_law_1.01,0.5691846466064453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,1,power_law_1.01,0.5962150573730469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,1,power_law_1.01,0.6319372940063477
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,8,1,power_law_1.01,12.26984619140625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,1,power_law_1.01,0.7309824371337891
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,1,power_law_1.01,0.7813241577148438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,1,power_law_1.01,0.9091289520263672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,1,power_law_1.01,1.0522534179687502
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,1,power_law_1.01,1.3215731811523437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,1,power_law_1.01,1.568975372314453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,1,power_law_1.01,2.101414337158203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,1,power_law_1.01,2.6591085815429687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,1,power_law_1.01,3.7401593017578123
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,1,power_law_1.01,4.781033020019531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,8,1,power_law_1.01,5.835056762695312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,1,power_law_1.2,0.07942463874816894
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,1,power_law_1.2,0.10748096466064454
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,1,power_law_1.2,0.18300672531127932
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,1,power_law_1.2,0.2671430397033691
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,1,power_law_1.2,0.34965694427490235
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,1,power_law_1.2,0.5730956649780273
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,1,power_law_1.2,0.6244780731201172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,1,power_law_1.2,0.6336959838867188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,1,power_law_1.2,0.6757523345947265
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,1,power_law_1.2,0.7042483520507813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,1,power_law_1.2,0.7484697723388672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,1,power_law_1.2,0.8038111877441405
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,1,power_law_1.2,0.8669651031494141
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,1,power_law_1.2,0.9708300781249999
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,1,power_law_1.2,1.0619110107421874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,1,power_law_1.2,1.1472281646728515
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,1,power_law_1.2,1.3668595886230468
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,1,power_law_1.2,1.0317068481445313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,1,power_law_1.2,1.158047332763672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,1,power_law_1.2,1.413923797607422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,1,power_law_1.2,1.8028370666503908
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,1,power_law_1.2,2.1505413818359376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,1,power_law_1.2,2.7902752685546877
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,1,power_law_1.2,3.6166439819335934
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,1,power_law_1.2,4.879486694335937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,1,power_law_1.2,6.7244152832031245
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,1,power_law_1.2,9.4840966796875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,1,power_law_1.2,0.054675841331481935
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,1,power_law_1.2,0.0910598373413086
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,1,power_law_1.2,0.19418304443359374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,1,power_law_1.2,0.12998463630676269
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,1,power_law_1.2,0.2710304069519043
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,1,power_law_1.2,0.43706497192382815
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,1,power_law_1.2,0.47009727478027347
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,1,power_law_1.2,0.4832793426513672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,1,power_law_1.2,0.5177382278442383
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,1,power_law_1.2,0.5054643249511719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,1,power_law_1.2,0.532058868408203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,1,power_law_1.2,0.5532908630371094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,1,power_law_1.2,0.5652659225463867
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,1,power_law_1.2,0.6016780853271484
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,1,power_law_1.2,0.598785285949707
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,1,power_law_1.2,0.6220332717895507
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,1,power_law_1.2,0.6662700653076172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,1,power_law_1.2,0.760755844116211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,1,power_law_1.2,0.8187942504882812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,1,power_law_1.2,0.9479039764404297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,1,power_law_1.2,1.0804550170898437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,1,power_law_1.2,1.3464282226562498
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,1,power_law_1.2,1.608020477294922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,1,power_law_1.2,2.149548797607422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,1,power_law_1.2,2.687818298339844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,1,power_law_1.2,3.7712466430664064
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,8,1,power_law_1.2,11.030428466796874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,1,power_law_1.2,4.841183471679687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,8,1,power_law_1.2,5.868630981445312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,1,power_law_1.01,0.04640960216522217
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,1,power_law_1.01,0.0525164794921875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,1,power_law_1.01,0.05528448104858399
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,1,power_law_1.01,0.06535615921020507
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,1,power_law_1.01,0.09338815689086914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,1,power_law_1.01,0.09558464050292968
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,1,power_law_1.01,0.09601728439331055
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,1,power_law_1.01,0.09692799568176269
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,1,power_law_1.01,0.09891263961791992
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,1,power_law_1.01,0.10100735664367674
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,1,power_law_1.01,0.10245247840881348
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,1,power_law_1.01,0.10531968116760253
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,1,power_law_1.01,0.10919679641723631
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,1,power_law_1.01,0.11235072135925292
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,1,power_law_1.01,0.1200556755065918
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,1,power_law_1.01,0.12469568252563476
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,1,power_law_1.01,0.14139583587646484
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,1,power_law_1.01,0.17736576080322267
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,1,power_law_1.01,0.2302022361755371
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,1,power_law_1.01,0.2780703926086426
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,1,power_law_1.01,0.37720321655273437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,1,power_law_1.01,0.47328254699707034
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,1,power_law_1.01,0.6722918701171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,1,power_law_1.01,0.870284194946289
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,1,power_law_1.01,1.2593209838867188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,1,power_law_1.01,0.032077438831329345
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,1,power_law_1.01,0.03492480039596558
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,1,power_law_1.01,0.03655551910400391
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,1,power_law_1.01,0.03984127998352051
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,1,power_law_1.01,0.04359231948852539
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,1,power_law_1.01,0.04543295860290527
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,1,power_law_1.01,0.04554880142211914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,1,power_law_1.01,0.04644224166870117
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,1,power_law_1.01,0.04669760227203369
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,1,power_law_1.01,0.047184000015258784
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,1,power_law_1.01,0.049304962158203125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,1,power_law_1.01,0.049200639724731446
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,1,power_law_1.01,0.05060991764068603
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,1,power_law_1.01,0.05674240112304687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,1,power_law_1.01,0.06095424175262452
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,1,power_law_1.01,0.06268223762512207
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,1,power_law_1.01,0.06458623886108399
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,1,power_law_1.01,0.0735206413269043
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,1,power_law_1.01,0.09039615631103516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,1,power_law_1.01,0.11361663818359376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,16,1,power_law_1.01,2.048966369628906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,1,power_law_1.01,0.12676608085632324
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,1,power_law_1.01,0.17296768188476563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,1,power_law_1.01,0.2233580780029297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,1,power_law_1.01,0.3047270393371582
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,1,power_law_1.01,0.4013459014892578
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,1,power_law_1.01,0.5747590255737305
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,1,power_law_1.01,0.7434413146972656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,16,1,power_law_1.01,0.019759360551834106
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,16,1,power_law_1.01,0.02125888109207153
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,16,1,power_law_1.01,0.018174079656600953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,16,1,power_law_1.01,0.02148736000061035
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,16,1,power_law_1.01,0.02060159921646118
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,16,1,power_law_1.01,0.021546239852905276
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,16,1,power_law_1.01,0.020594561100006105
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,16,1,power_law_1.01,0.021833600997924803
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,16,1,power_law_1.01,0.021457920074462893
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,16,1,power_law_1.01,0.021896960735321044
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,16,1,power_law_1.01,0.9054131317138673
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,16,1,power_law_1.01,0.02065216064453125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,16,1,power_law_1.01,0.024313600063323976
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,16,1,power_law_1.01,0.023327999114990235
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,16,1,power_law_1.01,0.025176320075988766
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,16,1,power_law_1.01,0.029436800479888912
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,16,1,power_law_1.01,0.03208319902420044
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,16,1,power_law_1.01,0.03550463914871216
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,16,1,power_law_1.01,0.04017087936401367
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.01,0.049764480590820316
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.01,0.06268095970153809
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.01,0.07533760070800781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.01,0.10012031555175782
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.01,0.12832639694213868
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.01,0.1833292770385742
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.01,0.23836544036865231
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,1,power_law_1.2,0.04689919948577881
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,1,power_law_1.2,0.05393343925476074
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,1,power_law_1.2,0.054779520034790044
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,1,power_law_1.2,0.06523392200469971
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,1,power_law_1.2,0.09359935760498048
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,1,power_law_1.2,0.09454272270202638
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,1,power_law_1.2,0.0959763240814209
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,1,power_law_1.2,0.09643391609191895
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,1,power_law_1.2,0.09815872192382813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,1,power_law_1.2,0.09925632476806641
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,1,power_law_1.2,0.10147135734558106
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,1,power_law_1.2,0.10276991844177245
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,1,power_law_1.2,0.10469568252563477
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,1,power_law_1.2,0.10932224273681641
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,1,power_law_1.2,0.11262911796569823
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,1,power_law_1.2,0.12204992294311523
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,1,power_law_1.2,0.12346048355102539
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,1,power_law_1.2,0.14157631874084473
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,1,power_law_1.2,0.17895679473876952
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.01,0.3501881790161133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,1,power_law_1.2,0.22990976333618166
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,1,power_law_1.2,0.2755660820007324
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,1,power_law_1.2,0.37695358276367186
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,1,power_law_1.2,0.47222080230712893
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.01,0.4674111938476562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,1,power_law_1.2,0.6688249969482423
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,16,1,power_law_1.01,0.5796582412719726
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,1,power_law_1.2,0.8608678436279297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,1,power_law_1.2,0.03422399997711182
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,1,power_law_1.2,0.03882047891616821
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,1,power_law_1.2,0.04116672039031982
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,1,power_law_1.2,0.0404256010055542
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,1,power_law_1.2,0.04442240238189697
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,1,power_law_1.2,0.04481664180755615
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,1,power_law_1.2,0.045969281196594235
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,1,power_law_1.2,0.04614719867706299
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,1,power_law_1.2,0.046869120597839355
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,1,power_law_1.2,0.047746558189392094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,1,power_law_1.2,0.048509440422058105
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,1,power_law_1.2,0.0509606409072876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,1,power_law_1.2,0.051928958892822265
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,1,power_law_1.2,0.058465919494628905
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,1,power_law_1.2,0.06238783836364746
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,1,power_law_1.2,0.06604032039642334
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,1,power_law_1.2,0.0669055986404419
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,1,power_law_1.2,0.07235455989837647
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,1,power_law_1.2,0.08975808143615722
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,1,power_law_1.2,0.11086272239685058
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,1,power_law_1.2,0.12604736328125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,1,power_law_1.2,1.2671334075927736
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,1,power_law_1.2,0.17244672775268555
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,1,power_law_1.2,0.22270591735839843
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,1,power_law_1.2,0.30254783630371096
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,1,power_law_1.2,1.669129638671875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,1,power_law_1.2,0.40144321441650394
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,16,1,power_law_1.2,2.0488812255859377
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,16,1,power_law_1.2,0.01901247978210449
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,16,1,power_law_1.2,0.019778560400009155
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,16,1,power_law_1.2,0.019809919595718383
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,16,1,power_law_1.2,0.021456639766693115
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,16,1,power_law_1.2,0.021376640796661378
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,16,1,power_law_1.2,0.02184448003768921
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,16,1,power_law_1.2,0.02147968053817749
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,16,1,power_law_1.2,0.021857919692993163
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,16,1,power_law_1.2,0.021838719844818118
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,16,1,power_law_1.2,0.021826560497283938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,16,1,power_law_1.2,0.022291839122772217
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,16,1,power_law_1.2,0.023486719131469727
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,16,1,power_law_1.2,0.025191679000854495
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,16,1,power_law_1.2,0.027148799896240233
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,16,1,power_law_1.2,0.029053440093994142
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,16,1,power_law_1.2,0.03332479953765869
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,1,power_law_1.2,0.5703680038452148
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,16,1,power_law_1.2,0.042712960243225094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,1,power_law_1.2,0.7423801422119141
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,16,1,power_law_1.2,0.03619263887405395
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.2,0.049491200447082516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.2,0.10424960136413575
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.2,0.0628223991394043
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.2,0.07735040187835693
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,16,1,power_law_1.2,0.914936294555664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.2,0.13184703826904295
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.2,0.1900422477722168
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.2,0.24776256561279295
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,1,power_law_1.01,0.06361599922180175
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,1,power_law_1.01,0.07190400123596191
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,1,power_law_1.01,0.07626368045806885
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,1,power_law_1.01,0.0898310375213623
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,1,power_law_1.01,0.12819647789001465
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,1,power_law_1.01,0.1296902370452881
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,1,power_law_1.01,0.13057151794433594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,1,power_law_1.01,0.13161855697631836
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,1,power_law_1.01,0.13266624450683592
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,1,power_law_1.01,0.13351936340332032
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,1,power_law_1.01,0.1395411205291748
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,1,power_law_1.01,0.1449612808227539
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,1,power_law_1.01,0.1501632022857666
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,1,power_law_1.01,0.15397184371948242
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,1,power_law_1.01,0.15855615615844726
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,1,power_law_1.01,0.1662067222595215
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,1,power_law_1.01,0.17566015243530272
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,1,power_law_1.01,0.22356096267700196
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,1,power_law_1.01,0.2926873588562012
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,1,power_law_1.01,0.3586252975463867
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,1,power_law_1.01,0.43349441528320315
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.2,0.3673196792602539
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,1,power_law_1.01,0.596635513305664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,1,power_law_1.01,0.7599481964111329
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,1,power_law_1.01,1.0758214569091797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,1,power_law_1.01,1.4228620910644532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,1,power_law_1.01,2.02923583984375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.2,0.48030910491943357
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,1,power_law_1.01,0.03837055921554565
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,1,power_law_1.01,0.04925695896148682
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,1,power_law_1.01,0.050532479286193845
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,1,power_law_1.01,0.06253503799438477
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,1,power_law_1.01,0.06752448081970215
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,1,power_law_1.01,0.06913472175598144
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,1,power_law_1.01,0.07040703773498536
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,1,power_law_1.01,0.07045631885528565
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,1,power_law_1.01,0.07063744068145753
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,1,power_law_1.01,0.07182784080505371
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,1,power_law_1.01,0.07275904178619384
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,1,power_law_1.01,0.07397503852844238
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,1,power_law_1.01,0.07596799850463867
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,1,power_law_1.01,0.0786572790145874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,1,power_law_1.01,2.864765930175781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,1,power_law_1.01,0.08203071594238281
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,1,power_law_1.01,0.08503295898437499
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,1,power_law_1.01,0.08730624198913575
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,1,power_law_1.01,0.09967616081237793
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,1,power_law_1.01,0.13104895591735838
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,1,power_law_1.01,0.15271360397338868
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,1,power_law_1.01,0.1762451171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,1,power_law_1.01,0.2476908874511719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,1,power_law_1.01,0.31124736785888674
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,1,power_law_1.01,0.4496697616577149
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,16,1,power_law_1.2,0.5955583953857422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,1,power_law_1.01,0.5664473724365234
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,16,1,power_law_1.01,3.42213134765625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,1,power_law_1.01,0.8385305786132813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,1,power_law_1.01,1.0793580627441406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,16,1,power_law_1.01,0.020430080890655518
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,16,1,power_law_1.01,0.028246400356292722
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,16,1,power_law_1.01,0.029743359088897702
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,16,1,power_law_1.01,0.03028671979904175
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,16,1,power_law_1.01,0.030623359680175783
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,16,1,power_law_1.01,0.02979327917098999
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,16,1,power_law_1.01,0.03006144046783447
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,16,1,power_law_1.01,0.031144959926605226
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,16,1,power_law_1.01,0.03176896095275879
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,16,1,power_law_1.01,0.023964800834655763
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,16,1,power_law_1.01,1.3351904296875001
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,16,1,power_law_1.01,0.023895680904388428
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,16,1,power_law_1.01,0.03187455892562866
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,16,1,power_law_1.01,0.0334278392791748
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,16,1,power_law_1.01,0.03940095901489258
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,16,1,power_law_1.01,0.04193535804748535
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,16,1,power_law_1.01,0.04334591865539551
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,16,1,power_law_1.01,0.04807744026184082
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.01,0.06782400131225585
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,16,1,power_law_1.01,0.05563968181610107
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.01,0.08506239891052246
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.01,0.10249216079711913
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.01,0.14060735702514648
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.01,0.18081024169921875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.01,0.25932159423828127
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.01,0.33855743408203126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,1,power_law_1.2,0.06408383846282958
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,1,power_law_1.2,0.07195199966430664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,1,power_law_1.2,0.07374144077301026
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,1,power_law_1.2,0.08864831924438477
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,1,power_law_1.2,0.12651264190673828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,1,power_law_1.2,0.12966272354125977
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,1,power_law_1.2,0.1311616039276123
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,1,power_law_1.2,0.131461763381958
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,1,power_law_1.2,0.1324300765991211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,1,power_law_1.2,0.1341049575805664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,1,power_law_1.2,0.13956031799316407
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,1,power_law_1.2,0.1466867160797119
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,1,power_law_1.2,0.15000831604003906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,1,power_law_1.2,0.15456000328063965
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,1,power_law_1.2,0.15940671920776367
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,1,power_law_1.2,0.16996864318847654
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,1,power_law_1.2,0.1770502471923828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,1,power_law_1.2,0.22477760314941406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,1,power_law_1.2,0.2976902389526367
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,1,power_law_1.2,0.3610585784912109
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.01,0.4992044830322266
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,1,power_law_1.2,0.43650367736816403
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,1,power_law_1.2,0.5987513732910157
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,1,power_law_1.2,0.7601849365234374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.01,0.6583679962158203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,1,power_law_1.2,1.074005126953125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,16,1,power_law_1.01,0.8244614410400392
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,1,power_law_1.2,1.411249237060547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,1,power_law_1.2,0.041483521461486816
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,1,power_law_1.2,0.04969024181365967
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,1,power_law_1.2,0.050343041419982915
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,1,power_law_1.2,0.062094078063964844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,1,power_law_1.2,0.06701824188232422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,1,power_law_1.2,0.06982783794403076
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,1,power_law_1.2,0.07057024002075195
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,1,power_law_1.2,0.07060287952423096
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,1,power_law_1.2,0.07191936016082764
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,1,power_law_1.2,0.07234240055084229
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,1,power_law_1.2,0.07343999862670898
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,1,power_law_1.2,0.0749132776260376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,1,power_law_1.2,0.07662079811096192
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,1,power_law_1.2,0.08005056381225586
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,1,power_law_1.2,0.08247615814208983
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,1,power_law_1.2,0.08549056053161622
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,1,power_law_1.2,0.09053376197814941
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,1,power_law_1.2,0.09944064140319823
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,1,power_law_1.2,0.13246591567993166
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,1,power_law_1.2,0.15290559768676756
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,1,power_law_1.2,0.17792703628540038
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,1,power_law_1.2,2.047315216064453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,1,power_law_1.2,0.2479046440124512
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,1,power_law_1.2,0.3108576011657715
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,1,power_law_1.2,2.803663330078125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,1,power_law_1.2,0.45076736450195315
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,1,power_law_1.2,0.5674585723876954
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,16,1,power_law_1.2,3.466211853027344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,16,1,power_law_1.2,0.021018879413604735
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,16,1,power_law_1.2,0.023903360366821287
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,16,1,power_law_1.2,0.02476288080215454
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,16,1,power_law_1.2,0.02880768060684204
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,16,1,power_law_1.2,0.029151360988616943
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,16,1,power_law_1.2,0.03002624034881592
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,16,1,power_law_1.2,0.03035264015197754
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,16,1,power_law_1.2,0.029960319995880124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,16,1,power_law_1.2,0.03064512014389038
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,16,1,power_law_1.2,0.030787200927734376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,16,1,power_law_1.2,0.03136255979537964
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,16,1,power_law_1.2,0.03256767988204956
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,16,1,power_law_1.2,0.03405312061309814
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,16,1,power_law_1.2,0.039903359413146974
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,16,1,power_law_1.2,0.04380608081817627
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,16,1,power_law_1.2,0.04501632213592529
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,1,power_law_1.2,0.8418656158447266
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,16,1,power_law_1.2,0.04872320175170898
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,16,1,power_law_1.2,0.05664447784423828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,1,power_law_1.2,1.0903392028808594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.2,0.10573439598083496
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.2,0.06894336223602296
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.2,0.08543295860290527
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,16,1,power_law_1.2,1.3343064880371094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.2,0.1476972770690918
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.2,0.18682239532470704
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.2,0.2668000030517578
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.2,0.34927486419677733
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,1,power_law_1.01,0.04520256042480469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,1,power_law_1.01,0.056723198890686034
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,1,power_law_1.01,0.0866483211517334
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,1,power_law_1.01,0.1108678436279297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,1,power_law_1.01,0.14158080101013185
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,1,power_law_1.01,0.18318592071533205
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,1,power_law_1.01,0.2382534408569336
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,1,power_law_1.01,0.2477177619934082
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,1,power_law_1.01,0.2509734344482422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,1,power_law_1.01,0.25528640747070314
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,1,power_law_1.01,0.2681612777709961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,1,power_law_1.01,0.29615039825439454
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,1,power_law_1.01,0.31633216857910157
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,1,power_law_1.01,0.3373875045776367
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,1,power_law_1.01,0.3615411376953125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,1,power_law_1.01,0.37050495147705076
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,1,power_law_1.01,0.3911648178100586
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,1,power_law_1.01,0.4393459320068359
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,1,power_law_1.01,0.5219699096679687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.2,0.5201011276245117
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,1,power_law_1.01,0.5893414306640625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,1,power_law_1.01,0.6567961883544922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,1,power_law_1.01,0.8225625610351562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,1,power_law_1.01,0.9948377227783203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,1,power_law_1.01,1.3291226196289063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,1,power_law_1.01,1.6793772888183593
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,1,power_law_1.01,2.34177978515625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,1,power_law_1.01,3.031292724609375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,1,power_law_1.01,0.04616960048675537
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,16,1,power_law_1.01,3.6957586669921874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,1,power_law_1.01,0.07313407897949219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,1,power_law_1.01,0.10520447731018065
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,1,power_law_1.01,0.052744321823120124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,1,power_law_1.01,0.14565312385559082
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,1,power_law_1.01,0.20115072250366212
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,1,power_law_1.01,0.25457984924316407
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,1,power_law_1.01,0.2640102386474609
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,1,power_law_1.01,0.26704896926879884
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,1,power_law_1.01,0.2721299171447754
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,1,power_law_1.01,0.27886592864990234
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,1,power_law_1.01,0.28883392333984376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,1,power_law_1.01,0.29161920547485354
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,1,power_law_1.01,0.30024768829345705
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,1,power_law_1.01,0.3287142562866211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,1,power_law_1.01,0.3357344055175781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,1,power_law_1.01,0.3470150375366211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,1,power_law_1.01,0.37030529022216796
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,1,power_law_1.01,0.3956185531616211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,1,power_law_1.01,0.4467071914672852
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,1,power_law_1.01,0.5109856033325195
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,1,power_law_1.01,0.6202687835693359
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,1,power_law_1.01,0.7210540771484375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,1,power_law_1.01,0.9477318572998048
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,1,power_law_1.01,1.1802207946777343
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.2,0.6887238311767578
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,1,power_law_1.01,1.641741485595703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,1,power_law_1.01,2.068714294433594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,16,1,power_law_1.01,2.534182434082031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,16,1,power_law_1.2,0.8485926055908204
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,16,1,power_law_1.01,0.068154878616333
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,16,1,power_law_1.01,0.11854399681091307
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,16,1,power_law_1.01,0.08363648414611816
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,16,1,power_law_1.01,0.13661248207092286
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,16,1,power_law_1.01,0.16937152862548827
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,16,1,power_law_1.01,0.17623231887817384
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,16,1,power_law_1.01,0.18095232009887696
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,16,1,power_law_1.01,0.18349056243896483
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,16,1,power_law_1.01,0.07103871822357177
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,16,1,power_law_1.01,0.19119295120239257
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,16,1,power_law_1.01,0.19668544769287108
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,16,1,power_law_1.01,0.20330688476562503
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,16,1,power_law_1.01,0.21159551620483397
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,16,1,power_law_1.01,0.2222879981994629
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,16,1,power_law_1.01,0.22998720169067383
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,16,1,power_law_1.01,0.2493209648132324
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,16,1,power_law_1.01,0.09684736251831054
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.01,0.3417766571044922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.01,0.4369804763793946
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,16,1,power_law_1.01,0.29228031158447265
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.01,0.5157766342163086
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.01,0.7060044860839844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.01,0.920904312133789
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.01,1.3105938720703125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.01,1.7339302062988282
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,1,power_law_1.2,0.04440447807312012
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,1,power_law_1.2,0.0686956787109375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,1,power_law_1.2,0.08752639770507813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,1,power_law_1.2,0.10981632232666017
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,1,power_law_1.2,0.14027584075927735
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,1,power_law_1.2,0.19163711547851564
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,1,power_law_1.2,0.2338060760498047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,1,power_law_1.2,0.24636159896850587
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,1,power_law_1.2,0.25047935485839845
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,1,power_law_1.2,0.25694847106933594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,1,power_law_1.2,0.268342399597168
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,1,power_law_1.2,0.30202112197875974
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,1,power_law_1.2,0.3204332733154297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,1,power_law_1.2,0.34401214599609375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,1,power_law_1.2,0.36503807067871097
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,1,power_law_1.2,0.37745662689208986
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,1,power_law_1.2,0.3983436965942383
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,1,power_law_1.2,0.4619475173950195
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,1,power_law_1.2,0.53840576171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,1,power_law_1.2,0.5951935958862304
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,1,power_law_1.2,0.6737446594238281
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,16,1,power_law_1.01,2.6862527465820314
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,1,power_law_1.2,0.8501030731201172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,1,power_law_1.2,1.0114854431152343
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,16,1,power_law_1.01,3.5002893066406253
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,1,power_law_1.2,1.3389907836914063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,1,power_law_1.2,1.6948281860351564
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,1,power_law_1.2,2.372386627197266
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,1,power_law_1.2,0.04607744216918945
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,1,power_law_1.2,3.036054382324219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,1,power_law_1.2,0.051762561798095706
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,1,power_law_1.2,0.07498239994049072
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,1,power_law_1.2,0.10052096366882324
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,1,power_law_1.2,0.14399104118347167
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,1,power_law_1.2,0.1960166358947754
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,1,power_law_1.2,0.25364479064941403
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,1,power_law_1.2,0.26507392883300784
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,1,power_law_1.2,0.2609676742553711
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,1,power_law_1.2,0.2713855934143067
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,7168,2048,8,256,16,1,power_law_1.01,4.304554748535156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,1,power_law_1.2,0.27939647674560547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,16,1,power_law_1.2,3.7465310668945313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,1,power_law_1.2,0.28620735168457034
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,1,power_law_1.2,0.3031635284423828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,1,power_law_1.2,0.2914080047607422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,1,power_law_1.2,0.3342201614379883
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,1,power_law_1.2,0.34077953338623046
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,1,power_law_1.2,0.3533241653442383
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,1,power_law_1.2,0.38256126403808594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,1,power_law_1.2,0.4101491165161133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,1,power_law_1.2,0.46591678619384763
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,1,power_law_1.2,0.5311142349243164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,1,power_law_1.2,0.6351443099975586
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,1,power_law_1.2,0.7627295684814454
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,1,power_law_1.2,0.9580127716064453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,1,power_law_1.2,1.2009625244140625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,1,power_law_1.2,1.6495263671875002
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,16,1,power_law_1.2,0.06848320007324218
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,1,power_law_1.2,2.1143399047851563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,16,1,power_law_1.2,0.07147071838378907
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,16,1,power_law_1.2,0.08203136444091796
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,16,1,power_law_1.2,2.5373606872558594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,16,1,power_law_1.2,0.09506752014160155
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,16,1,power_law_1.2,0.11358528137207032
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,16,1,power_law_1.2,0.1324115180969238
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,16,1,power_law_1.2,0.16757312774658203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,16,1,power_law_1.2,0.17498943328857422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,16,1,power_law_1.2,0.17939199447631834
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,16,1,power_law_1.2,0.18514432907104492
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,16,1,power_law_1.2,0.1919647979736328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,16,1,power_law_1.2,0.2006028747558594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,16,1,power_law_1.2,0.2068320083618164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,16,1,power_law_1.2,0.22311935424804688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,16,1,power_law_1.2,0.234401912689209
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,16,1,power_law_1.2,0.24767232894897462
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,16,1,power_law_1.2,0.27673408508300784
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,16,1,power_law_1.2,0.34086719512939456
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.2,0.4107328033447265
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.2,0.5540345764160156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.2,0.6836544036865234
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.2,1.028808288574219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.2,1.3740850830078126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.2,2.0515032958984376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.2,2.793204345703125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,1,power_law_1.01,0.0665388822555542
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,1,power_law_1.01,0.06540351867675781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,1,power_law_1.01,0.0767033576965332
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,1,power_law_1.01,0.09232831954956054
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,1,power_law_1.01,0.09617152214050292
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,1,power_law_1.01,0.09779135704040527
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,1,power_law_1.01,0.10401151657104493
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,1,power_law_1.01,0.1111577606201172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,1,power_law_1.01,0.14192831993103028
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,1,power_law_1.01,0.14418944358825683
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,1,power_law_1.01,0.1498905563354492
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,1,power_law_1.01,0.15274432182312012
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,1,power_law_1.01,0.15593024253845217
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,1,power_law_1.01,0.1646188735961914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,1,power_law_1.01,0.1833683204650879
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,1,power_law_1.01,0.19998271942138673
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,1,power_law_1.01,0.23717695236206054
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,1,power_law_1.01,0.28153600692749026
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,1,power_law_1.01,0.39316673278808595
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,1,power_law_1.01,0.4723199844360352
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,1,power_law_1.01,0.6628428649902344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,1,power_law_1.01,0.8394092559814453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,1,power_law_1.01,1.202590103149414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,1,power_law_1.01,1.5612326049804688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,16,1,power_law_1.01,1.9264378356933594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,1,power_law_1.01,0.03628096103668213
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,1,power_law_1.01,0.04034815788269043
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,1,power_law_1.01,0.052876801490783686
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,1,power_law_1.01,0.06211264133453369
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,1,power_law_1.01,0.07665472030639649
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,1,power_law_1.01,0.09813119888305663
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,1,power_law_1.01,0.09918527603149414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,1,power_law_1.01,0.10399871826171876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,1,power_law_1.01,0.10392255783081054
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,1,power_law_1.01,0.10791040420532227
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,1,power_law_1.01,0.10777215957641602
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,1,power_law_1.01,0.1238924789428711
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,1,power_law_1.01,0.12268992424011231
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,1,power_law_1.01,0.13174464225769042
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,1,power_law_1.01,0.12936320304870605
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,1,power_law_1.01,0.13196096420288086
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,1,power_law_1.01,0.14069248199462892
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,1,power_law_1.01,0.1537887954711914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,1,power_law_1.01,0.17888639450073243
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,1,power_law_1.01,0.2115212821960449
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,1,power_law_1.01,0.24597375869750976
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,1,power_law_1.01,0.3096614456176758
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,1,power_law_1.01,0.38493824005126953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,1,power_law_1.01,0.5261708831787109
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,1,power_law_1.01,0.6538355255126953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,1,power_law_1.01,0.934427490234375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,1,power_law_1.01,1.2123283386230468
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,16,1,power_law_1.01,1.4850968933105468
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,16,1,power_law_1.01,0.04517824172973633
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,16,1,power_law_1.01,0.05774335861206055
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,16,1,power_law_1.01,0.06372863769531249
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,16,1,power_law_1.01,0.07992127895355225
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,16,1,power_law_1.01,0.08382592201232911
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,16,1,power_law_1.01,0.08578047752380372
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,16,1,power_law_1.01,0.08679360389709473
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,16,1,power_law_1.01,0.09195008277893066
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,16,1,power_law_1.01,0.0941983985900879
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,16,1,power_law_1.01,0.09677696228027344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,16,1,power_law_1.01,0.10296256065368652
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,16,1,power_law_1.01,0.04642303943634033
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,16,1,power_law_1.01,0.050241279602050784
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,16,1,power_law_1.01,0.08629631996154785
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,16,1,power_law_1.01,0.11378687858581542
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,16,1,power_law_1.01,0.11493824005126954
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,16,1,power_law_1.01,0.12656448364257813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,16,1,power_law_1.01,0.15173376083374024
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,16,1,power_law_1.01,0.18230464935302734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,16,1,power_law_1.01,0.23713920593261717
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,16,1,power_law_1.01,0.29824256896972656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,16,1,power_law_1.01,0.419317741394043
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,16,1,power_law_1.01,0.5469830322265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,16,1,power_law_1.01,0.8012960052490234
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,16,1,power_law_1.01,1.064673309326172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,1,power_law_1.2,0.03656192064285278
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,16,1,power_law_1.01,1.639411163330078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,1,power_law_1.2,0.051357440948486335
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,1,power_law_1.2,0.06562880039215088
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,1,power_law_1.2,0.07874623775482179
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,1,power_law_1.2,0.10006527900695801
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,1,power_law_1.2,0.09558591842651368
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,1,power_law_1.2,0.10272255897521973
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,1,power_law_1.2,0.10534399986267089
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,1,power_law_1.2,0.11140288352966307
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,1,power_law_1.2,0.1417632007598877
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,1,power_law_1.2,0.14460415840148927
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,1,power_law_1.2,0.057299838066101075
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,1,power_law_1.2,0.14956288337707518
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,1,power_law_1.2,0.1521440029144287
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,1,power_law_1.2,0.15603584289550781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,1,power_law_1.2,0.16497600555419922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,1,power_law_1.2,0.18511232376098632
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,1,power_law_1.2,0.2008569526672363
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,1,power_law_1.2,0.23936384201049804
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,1,power_law_1.2,0.2880748748779297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,16,1,power_law_1.01,2.759747314453125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,1,power_law_1.2,0.3958502578735351
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,1,power_law_1.2,0.47356929779052737
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,1,power_law_1.2,0.6688339233398437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,1,power_law_1.2,0.8490239715576171
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,1,power_law_1.2,1.2114476776123047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,1,power_law_1.2,0.035968000888824465
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,1,power_law_1.2,0.03986560106277466
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,1,power_law_1.2,0.052312321662902825
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,1,power_law_1.2,0.06160575866699218
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,1,power_law_1.2,0.07727935791015625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,1,power_law_1.2,0.10300864219665526
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,1,power_law_1.2,0.10087039947509766
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,1,power_law_1.2,0.10127167701721192
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,1,power_law_1.2,0.10638976097106934
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,1,power_law_1.2,0.10479743957519531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,1,power_law_1.2,0.10902272224426271
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,1,power_law_1.2,0.12264575958251953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,1,power_law_1.2,0.12770496368408205
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,1,power_law_1.2,0.12895872116088866
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,1,power_law_1.2,0.13119039535522461
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,1,power_law_1.2,0.1338931179046631
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,1,power_law_1.2,0.14191679954528807
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,1,power_law_1.2,0.15745535850524903
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,1,power_law_1.2,0.1833395195007324
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,1,power_law_1.2,0.2167967987060547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,1,power_law_1.2,0.24966527938842775
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,1,power_law_1.2,0.3107967948913574
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,1,power_law_1.2,0.3883871841430664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,1,power_law_1.2,0.5287020874023438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,1,power_law_1.2,1.5805325317382812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,1,power_law_1.2,0.6498470306396484
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,1,power_law_1.2,0.9354656219482422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,16,1,power_law_1.2,0.045135998725891115
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,16,1,power_law_1.2,0.04691135883331299
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,16,1,power_law_1.2,0.05142975807189941
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,16,1,power_law_1.2,0.055738239288330084
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,16,1,power_law_1.2,0.06192192077636719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,16,1,power_law_1.2,0.08006400108337403
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,16,1,power_law_1.2,0.0834598445892334
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,16,1,power_law_1.2,0.08560895919799805
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,16,1,power_law_1.2,0.08730879783630371
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,16,1,power_law_1.2,0.0888441562652588
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,16,1,power_law_1.2,0.092291841506958
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,16,1,power_law_1.2,0.09585280418395996
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,16,1,power_law_1.2,0.09862591743469237
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,16,1,power_law_1.2,0.11237631797790529
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,16,1,power_law_1.2,0.11898624420166015
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,16,1,power_law_1.2,0.1269868755340576
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,16,1,power_law_1.2,1.9455877685546874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,16,1,power_law_1.2,0.14293312072753905
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,16,1,power_law_1.2,0.18039039611816404
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,1,power_law_1.2,1.2164051055908203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,16,1,power_law_1.2,0.21936767578125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,16,1,power_law_1.2,0.3071001625061035
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,16,1,power_law_1.2,0.3901331329345703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,16,1,power_law_1.2,1.4818194580078123
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,16,1,power_law_1.2,0.574263038635254
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,16,1,power_law_1.2,0.7689126586914062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,16,1,power_law_1.2,1.1372576141357422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,16,1,power_law_1.2,1.4950047302246094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,1,power_law_1.01,0.04476607799530029
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,1,power_law_1.01,0.07050559997558595
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,1,power_law_1.01,0.08838591575622559
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,1,power_law_1.01,0.10800064086914063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,1,power_law_1.01,0.13340288162231445
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,1,power_law_1.01,0.19219776153564455
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,1,power_law_1.01,0.18898303985595705
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,1,power_law_1.01,0.19531007766723635
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,1,power_law_1.01,0.20214464187622072
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,1,power_law_1.01,0.20951679229736325
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,1,power_law_1.01,0.2246892738342285
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,1,power_law_1.01,0.24609983444213865
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,1,power_law_1.01,0.24799808502197265
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,1,power_law_1.01,0.2577907180786133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,1,power_law_1.01,0.2622457695007324
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,1,power_law_1.01,0.2672960090637207
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,1,power_law_1.01,0.28545728683471683
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,1,power_law_1.01,0.3347942352294922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,1,power_law_1.01,0.4038502502441406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,1,power_law_1.01,0.47566848754882807
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,1,power_law_1.01,0.554013442993164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,1,power_law_1.01,0.7283071899414063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,16,1,power_law_1.2,2.2521478271484376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,1,power_law_1.01,0.8782637023925781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,1,power_law_1.01,1.2147277069091795
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,1,power_law_1.01,1.5410028076171876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,1,power_law_1.01,2.2029132080078124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,1,power_law_1.01,2.860199584960937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,1,power_law_1.01,0.04719808101654053
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,1,power_law_1.01,0.05297664165496826
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,1,power_law_1.01,0.07446080207824707
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,16,1,power_law_1.01,3.52383544921875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,1,power_law_1.01,0.09733504295349121
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,1,power_law_1.01,0.19265024185180663
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,1,power_law_1.01,0.12462847709655762
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,1,power_law_1.01,0.18786560058593751
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,1,power_law_1.01,0.19222143173217773
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,1,power_law_1.01,0.19607616424560548
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,1,power_law_1.01,0.19893247604370118
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,1,power_law_1.01,0.20297216415405273
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,1,power_law_1.01,0.2093337631225586
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,1,power_law_1.01,0.24311807632446286
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,1,power_law_1.01,0.25764543533325196
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,1,power_law_1.01,0.2589331245422363
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,1,power_law_1.01,0.24878911972045897
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,1,power_law_1.01,0.28232831954956056
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,1,power_law_1.01,0.30490495681762697
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,1,power_law_1.01,0.32602558135986326
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,1,power_law_1.01,0.3755238342285156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,1,power_law_1.01,0.4515532684326172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,1,power_law_1.01,0.5675187301635742
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,1,power_law_1.01,0.6767040252685547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,1,power_law_1.01,0.9259481811523438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,1,power_law_1.01,1.1609907531738282
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,1,power_law_1.01,1.625964813232422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,1,power_law_1.01,2.098448028564453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,16,1,power_law_1.01,2.5740780639648437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,16,1,power_law_1.2,2.9938009643554686
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,16,1,power_law_1.01,0.05132351875305176
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,16,1,power_law_1.01,0.07877567768096924
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,16,1,power_law_1.01,0.055774722099304196
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,16,1,power_law_1.01,0.09151103973388672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,16,1,power_law_1.01,0.12438079833984375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,16,1,power_law_1.01,0.13041024208068847
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,16,1,power_law_1.01,0.1345248031616211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,16,1,power_law_1.01,0.13965184211730958
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,16,1,power_law_1.01,0.1410745620727539
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,16,1,power_law_1.01,0.14727359771728515
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,16,1,power_law_1.01,0.15251839637756348
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,16,1,power_law_1.01,0.15496319770812989
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,16,1,power_law_1.01,0.16592639923095703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,16,1,power_law_1.01,0.1724140739440918
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,16,1,power_law_1.01,0.17632959365844728
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,16,1,power_law_1.01,0.0668671989440918
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,16,1,power_law_1.2,3.7335769653320314
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,16,1,power_law_1.01,0.1984160041809082
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,16,1,power_law_1.01,0.23740224838256835
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,16,1,power_law_1.01,0.27206464767456057
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,16,1,power_law_1.01,0.35421569824218746
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,16,1,power_law_1.01,0.435346565246582
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,16,1,power_law_1.01,0.6078899383544922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,16,1,power_law_1.01,0.7893619537353516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,16,1,power_law_1.01,1.5280210876464844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,1,power_law_1.2,0.042501120567321776
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,1,power_law_1.2,0.07183807849884033
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,16,1,power_law_1.01,2.315619812011719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,16,1,power_law_1.01,3.118358459472656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,1,power_law_1.2,0.08740927696228026
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,1,power_law_1.2,0.10084927558898926
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,1,power_law_1.2,0.19384895324707033
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,1,power_law_1.2,0.18799232482910155
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,1,power_law_1.2,0.20083967208862305
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,1,power_law_1.2,0.12952192306518556
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,1,power_law_1.2,0.20908159255981446
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,1,power_law_1.2,0.22057024002075196
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,1,power_law_1.2,0.26180032730102537
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,1,power_law_1.2,0.25446271896362305
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,1,power_law_1.2,0.2676019287109375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,1,power_law_1.2,0.26086912155151365
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,1,power_law_1.2,0.2739795112609863
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,1,power_law_1.2,0.2929593658447266
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,1,power_law_1.2,0.23122047424316405
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,1,power_law_1.2,0.3453638458251953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,1,power_law_1.2,0.4229228973388672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,1,power_law_1.2,0.48133121490478514
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,1,power_law_1.2,0.5586240005493164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,1,power_law_1.2,0.7339398193359374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,1,power_law_1.2,0.891520004272461
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,1,power_law_1.2,1.218757095336914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,1,power_law_1.2,1.5496006774902344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,1,power_law_1.2,2.2239007568359375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,1,power_law_1.2,0.046451201438903814
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,1,power_law_1.2,0.05956031799316406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,1,power_law_1.2,0.07333759784698486
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,1,power_law_1.2,0.09344639778137206
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,1,power_law_1.2,0.1259404754638672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,1,power_law_1.2,0.19252992630004884
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,1,power_law_1.2,0.1878771209716797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,1,power_law_1.2,0.19333503723144532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,1,power_law_1.2,0.19929855346679687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,1,power_law_1.2,0.20213951110839842
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,1,power_law_1.2,0.20776384353637695
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,1,power_law_1.2,0.21188352584838865
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,1,power_law_1.2,0.26111999511718753
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,1,power_law_1.2,0.26145727157592774
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,1,power_law_1.2,0.27242111206054687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,1,power_law_1.2,2.871898193359375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,1,power_law_1.2,0.28987712860107423
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,1,power_law_1.2,0.279486083984375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,1,power_law_1.2,0.308221435546875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,1,power_law_1.2,0.33200511932373045
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,1,power_law_1.2,0.3860575866699219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,1,power_law_1.2,0.4568415832519531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,1,power_law_1.2,0.586255989074707
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,1,power_law_1.2,0.6876255798339843
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,1,power_law_1.2,0.9366111755371094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,1,power_law_1.2,1.1660102081298827
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,16,1,power_law_1.2,3.537353515625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,1,power_law_1.2,1.6486470031738283
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,16,1,power_law_1.2,0.05192063808441162
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,16,1,power_law_1.2,0.0651417589187622
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,16,1,power_law_1.2,0.07396416187286377
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,16,1,power_law_1.2,0.08866559982299804
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,16,1,power_law_1.2,0.12506175994873048
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,16,1,power_law_1.2,0.1318489646911621
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,16,1,power_law_1.2,0.1385747241973877
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,16,1,power_law_1.2,0.14455039978027345
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,1,power_law_1.2,2.1140115356445315
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,16,1,power_law_1.2,0.148089599609375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,16,1,power_law_1.2,0.15548928260803224
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,16,1,power_law_1.2,0.15787839889526367
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,16,1,power_law_1.2,0.16637184143066405
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,16,1,power_law_1.2,0.17628992080688477
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,16,1,power_law_1.2,0.18826879501342772
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,16,1,power_law_1.2,0.05532800197601319
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,16,1,power_law_1.2,2.582641906738281
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,16,1,power_law_1.2,0.2020102310180664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,16,1,power_law_1.2,0.22474431991577148
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,16,1,power_law_1.2,0.2751296043395996
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,16,1,power_law_1.2,0.33495361328125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,16,1,power_law_1.2,0.5958067321777344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,16,1,power_law_1.2,0.4624185562133789
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,16,1,power_law_1.2,0.8959366607666016
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,16,1,power_law_1.2,1.2080915069580078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,16,1,power_law_1.2,1.8689517211914062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,16,1,power_law_1.2,2.5558848571777344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,1,power_law_1.01,0.04862080097198486
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,1,power_law_1.01,0.06984576225280761
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,1,power_law_1.01,0.09123071670532226
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,1,power_law_1.01,0.11738752365112304
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,1,power_law_1.01,0.16532543182373047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,1,power_law_1.01,0.24032703399658203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,1,power_law_1.01,0.26441791534423825
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,1,power_law_1.01,0.3378521728515625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,1,power_law_1.01,0.35189247131347656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,1,power_law_1.01,0.36407295227050784
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,1,power_law_1.01,0.37507583618164064
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,1,power_law_1.01,0.38290431976318356
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,1,power_law_1.01,0.39897598266601564
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,1,power_law_1.01,0.41439678192138674
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,1,power_law_1.01,0.43160385131835943
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,1,power_law_1.01,0.4610918426513672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,1,power_law_1.01,0.5920563125610352
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,1,power_law_1.01,0.6150688171386719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,1,power_law_1.01,0.6456556701660157
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,1,power_law_1.01,0.7701824188232422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,1,power_law_1.01,0.8496761322021484
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,16,1,power_law_1.2,3.9325747680664067
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,1,power_law_1.01,0.9821644592285157
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,1,power_law_1.01,1.5100274658203126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,1,power_law_1.01,1.1529984283447265
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,1,power_law_1.01,1.8498899841308591
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,1,power_law_1.01,2.554562530517578
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,1,power_law_1.01,3.2518585205078123
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,1,power_law_1.01,0.055577597618103027
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,1,power_law_1.01,0.07187903881072998
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,1,power_law_1.01,0.11094079971313478
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,1,power_law_1.01,0.16922176361083985
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,16,1,power_law_1.01,3.960015869140625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,1,power_law_1.01,0.26102848052978517
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,1,power_law_1.01,0.2959212875366211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,1,power_law_1.01,0.3376038360595703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,1,power_law_1.01,0.3483206558227539
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,1,power_law_1.01,0.36032127380371093
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,1,power_law_1.01,0.3692364883422852
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,1,power_law_1.01,0.38863296508789064
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,16,1,power_law_1.2,5.3016290283203125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,1,power_law_1.01,0.37574718475341795
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,1,power_law_1.01,0.40642623901367186
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,1,power_law_1.01,0.40690048217773434
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,1,power_law_1.01,0.42540351867675785
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,1,power_law_1.01,0.47531967163085936
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,1,power_law_1.01,0.4951193618774414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,1,power_law_1.01,0.5230931091308594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,1,power_law_1.01,0.5774860763549805
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,1,power_law_1.01,0.6291910552978516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,1,power_law_1.01,0.7523628997802734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,1,power_law_1.01,0.8738253021240234
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,1,power_law_1.01,1.114092788696289
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,1,power_law_1.01,1.3506285095214845
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,1,power_law_1.01,1.8219206237792966
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,1,power_law_1.01,2.2887557983398437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,16,1,power_law_1.01,2.7844326782226565
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,16,1,power_law_1.01,0.08411392211914062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,16,1,power_law_1.01,0.08444160461425781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,16,1,power_law_1.01,0.08620223999023438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,16,1,power_law_1.01,0.08938752174377441
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,16,1,power_law_1.01,0.09623871803283691
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,16,1,power_law_1.01,0.10335488319396972
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,16,1,power_law_1.01,0.08530688285827637
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,16,1,power_law_1.01,0.10982080459594727
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,16,1,power_law_1.01,0.1163865566253662
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,16,1,power_law_1.01,0.12263360023498535
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,16,1,power_law_1.01,0.13691072463989257
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,16,1,power_law_1.01,0.16439872741699219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,16,1,power_law_1.01,0.22485183715820312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,16,1,power_law_1.01,0.19279552459716798
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,16,1,power_law_1.01,0.29160255432128906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,16,1,power_law_1.01,0.3319071960449219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,16,1,power_law_1.01,0.5947423934936523
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,16,1,power_law_1.01,0.42108417510986323
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,16,1,power_law_1.01,0.7717772674560547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,16,1,power_law_1.01,1.1256915283203124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,16,1,power_law_1.01,1.469063720703125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,16,1,power_law_1.01,2.183391418457031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,16,1,power_law_1.01,2.9151885986328123
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,1,power_law_1.2,0.04728127956390381
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,1,power_law_1.2,0.059873919486999515
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,16,1,power_law_1.01,4.332574157714843
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,1,power_law_1.2,0.09027008056640626
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,1,power_law_1.2,0.1277996826171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,1,power_law_1.2,0.15409728050231936
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,1,power_law_1.2,0.21342399597167966
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,1,power_law_1.2,0.2511104011535644
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,1,power_law_1.2,0.3321420669555664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,1,power_law_1.2,0.3447052764892578
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,1,power_law_1.2,0.3532793426513672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,1,power_law_1.2,0.3681561660766602
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,1,power_law_1.2,0.3795251083374024
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,1,power_law_1.2,0.3935539245605469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,1,power_law_1.2,0.41047039031982424
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,1,power_law_1.2,0.4367808151245117
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,1,power_law_1.2,0.46622528076171876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,1,power_law_1.2,0.5887507247924805
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,1,power_law_1.2,0.6118604660034179
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,1,power_law_1.2,0.6468109130859375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,1,power_law_1.2,0.7815897369384766
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,1,power_law_1.2,0.8553459167480468
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,1,power_law_1.2,0.9863648223876954
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,1,power_law_1.2,1.168760986328125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,1,power_law_1.2,1.5146771240234376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,1,power_law_1.2,1.874113311767578
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,1,power_law_1.2,2.5592568969726566
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,1,power_law_1.2,3.2800228881835936
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,1,power_law_1.2,0.04935679912567138
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,1,power_law_1.2,0.055002880096435544
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,1,power_law_1.2,0.0763481616973877
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,1,power_law_1.2,0.10902272224426271
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,16,1,power_law_1.2,3.961611633300781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,1,power_law_1.2,0.15598848342895508
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,1,power_law_1.2,0.2274118423461914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,1,power_law_1.2,0.28493824005126955
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,1,power_law_1.2,0.33230976104736326
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,1,power_law_1.2,0.3425376129150391
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,1,power_law_1.2,0.3488889694213867
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,1,power_law_1.2,0.3607360076904297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,1,power_law_1.2,0.371960334777832
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,1,power_law_1.2,0.3849497604370117
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,1,power_law_1.2,0.40005119323730465
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,1,power_law_1.2,0.4088390350341797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,1,power_law_1.2,0.42118846893310546
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,1,power_law_1.2,0.47297023773193364
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,1,power_law_1.2,0.5036735916137696
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,1,power_law_1.2,0.5322323226928711
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,1,power_law_1.2,0.5891737747192383
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,1,power_law_1.2,0.6522489929199219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,1,power_law_1.2,0.7858322906494141
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,1,power_law_1.2,0.8998162841796875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,1,power_law_1.2,1.1378233337402344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,1,power_law_1.2,1.3826521301269532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,1,power_law_1.2,1.8488255310058594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,1,power_law_1.2,2.3261459350585936
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,16,1,power_law_1.2,0.08403136253356933
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,16,1,power_law_1.2,2.7984939575195313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,16,1,power_law_1.2,0.08412991523742676
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,16,1,power_law_1.2,0.08526464462280273
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,16,1,power_law_1.2,0.08966783523559571
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,16,1,power_law_1.2,0.09644607543945313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,16,1,power_law_1.2,0.1025817584991455
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,16,1,power_law_1.2,0.10989184379577636
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,16,1,power_law_1.2,0.11680383682250976
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,16,1,power_law_1.2,0.12349760055541992
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,16,1,power_law_1.2,0.1365171241760254
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,16,1,power_law_1.2,0.17144639968872072
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,16,1,power_law_1.2,0.18291519165039063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,16,1,power_law_1.2,0.22822399139404298
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,16,1,power_law_1.2,0.2912812805175781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,16,1,power_law_1.2,0.08502335548400879
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,16,1,power_law_1.2,0.42335231781005855
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,16,1,power_law_1.2,0.5981856155395507
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,16,1,power_law_1.2,0.3330489730834961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,16,1,power_law_1.2,1.1260128021240234
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,16,1,power_law_1.2,0.7689401245117187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,16,1,power_law_1.2,1.4750079345703124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,16,1,power_law_1.2,2.173317718505859
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,16,1,power_law_1.2,2.9139865112304686
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,16,1,power_law_1.2,4.323914794921874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,1,power_law_1.01,0.060312318801879886
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,1,power_law_1.01,0.06770559787750244
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,1,power_law_1.01,0.08133055686950683
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,1,power_law_1.01,0.11120960235595703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,1,power_law_1.01,0.11711744308471679
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,1,power_law_1.01,0.12130687713623048
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,1,power_law_1.01,0.12574399948120116
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,1,power_law_1.01,0.13036864280700683
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,1,power_law_1.01,0.13964863777160647
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,1,power_law_1.01,0.168024959564209
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,1,power_law_1.01,0.1828838348388672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,1,power_law_1.01,0.1932806396484375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,1,power_law_1.01,0.19450624465942384
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,1,power_law_1.01,0.19863807678222656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,1,power_law_1.01,0.20829696655273439
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,1,power_law_1.01,0.22870527267456056
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,1,power_law_1.01,0.24993919372558593
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,1,power_law_1.01,0.3067136001586914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,1,power_law_1.01,0.3564992141723633
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,1,power_law_1.01,0.4580403137207031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,1,power_law_1.01,0.5645536041259765
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,1,power_law_1.01,0.7593222045898438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,1,power_law_1.01,0.9734092712402344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,1,power_law_1.01,1.3805920410156252
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,1,power_law_1.01,1.7946617126464843
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,16,1,power_law_1.01,2.2134080505371094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,1,power_law_1.2,0.030161919593811037
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,1,power_law_1.2,0.05315904140472412
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,1,power_law_1.2,0.0636627197265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,1,power_law_1.2,0.07760704040527344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,1,power_law_1.2,0.11143296241760255
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,1,power_law_1.2,0.11843456268310545
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,1,power_law_1.2,0.12329216003417968
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,1,power_law_1.2,0.12756735801696778
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,1,power_law_1.2,0.13302271842956542
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,1,power_law_1.2,0.14164159774780274
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,1,power_law_1.2,0.18807104110717773
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,1,power_law_1.2,0.19092479705810547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,1,power_law_1.2,0.19666624069213867
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,1,power_law_1.2,0.19780031204223633
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,1,power_law_1.2,0.20344768524169923
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,1,power_law_1.2,0.2119430351257324
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,1,power_law_1.2,0.03459712028503418
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,1,power_law_1.2,0.23192256927490235
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,1,power_law_1.2,0.3180761528015137
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,1,power_law_1.2,0.3581760025024414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,1,power_law_1.2,0.24936895370483397
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,1,power_law_1.2,0.46084735870361326
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,1,power_law_1.2,0.5598803329467773
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,1,power_law_1.2,0.7572480010986328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,1,power_law_1.2,0.976346206665039
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,1,power_law_1.2,1.3830239868164063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,1,power_law_1.01,0.07234176158905029
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,1,power_law_1.01,0.17903615951538085
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,1,power_law_1.01,0.11585344314575194
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,1,power_law_1.01,0.41977985382080074
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,1,power_law_1.01,0.6515385437011718
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,1,power_law_1.01,0.2877689552307129
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,1,power_law_1.01,0.6984774780273437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,1,power_law_1.2,1.8092646789550781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,1,power_law_1.01,0.7525421142578125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,1,power_law_1.01,0.7265471649169922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,1,power_law_1.01,0.837523193359375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,1,power_law_1.01,0.9072191619873047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,1,power_law_1.01,0.9650649261474609
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,1,power_law_1.01,0.7759603118896484
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,1,power_law_1.01,1.1130528259277344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,1,power_law_1.01,1.1987999725341798
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,1,power_law_1.01,1.3373529052734374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,1,power_law_1.01,1.5830610656738282
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,1,power_law_1.01,1.2561984252929688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,1,power_law_1.01,1.3992063903808594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,1,power_law_1.01,2.0303007507324216
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,1,power_law_1.01,2.6658035278320313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,1,power_law_1.01,1.7228703308105469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,16,1,power_law_1.2,2.2268453979492184
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,1,power_law_1.01,3.2479833984375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,1,power_law_1.01,4.517592468261719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,1,power_law_1.01,5.762503051757813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,1,power_law_1.01,8.229912109375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,1,power_law_1.01,10.712816162109375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,16,1,power_law_1.01,13.304689941406249
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,1,power_law_1.01,0.24168703079223636
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,1,power_law_1.01,0.37774593353271485
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,1,power_law_1.01,0.6311756896972656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,1,power_law_1.01,1.0715264129638673
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,1,power_law_1.01,1.472265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,1,power_law_1.01,2.9689370727539064
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,1,power_law_1.01,2.6859512329101562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,1,power_law_1.01,2.963321533203125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,1,power_law_1.01,3.4413958740234376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,1,power_law_1.01,3.8549645996093753
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,1,power_law_1.01,2.5155097961425783
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,1,power_law_1.01,3.5930911254882814
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,1,power_law_1.01,3.078771057128906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,1,power_law_1.01,2.8328018188476562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,1,power_law_1.01,4.308631591796875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,1,power_law_1.01,4.091455993652344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,1,power_law_1.01,3.6032952880859375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,1,power_law_1.01,5.435075073242187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,1,power_law_1.01,5.322857055664063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,1,power_law_1.01,5.862893676757812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,1,power_law_1.01,6.3331982421875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,1,power_law_1.01,9.549673461914063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,1,power_law_1.01,10.294830322265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,1,power_law_1.01,13.041563720703124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,1,power_law_1.01,16.022550048828126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,1,power_law_1.2,0.07796288013458252
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,1,power_law_1.2,0.11695103645324707
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,1,power_law_1.2,0.17952320098876953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,1,power_law_1.2,0.2708211135864258
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,1,power_law_1.2,0.3809952163696289
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,1,power_law_1.2,0.6431423950195312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,1,power_law_1.2,0.6996063995361328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,1,power_law_1.2,0.7197203063964844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,1,power_law_1.2,0.7594815826416015
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,1,power_law_1.2,0.7979135894775391
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,1,power_law_1.2,0.8596966552734375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,1,power_law_1.2,0.936407699584961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,1,power_law_1.2,1.000913314819336
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,1,power_law_1.01,26.6062109375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,1,power_law_1.2,1.1414342498779297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,1,power_law_1.2,1.2551904296875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,1,power_law_1.2,1.3657606506347657
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,1,power_law_1.2,1.6231890869140624
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,1,power_law_1.2,1.295477752685547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,1,power_law_1.2,1.4459423828125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,1,power_law_1.01,31.10338623046875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,1,power_law_1.2,1.748389129638672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,1,power_law_1.2,2.069964752197266
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,1,power_law_1.2,2.6890661621093748
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,1,power_law_1.2,3.2797445678710937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,16,1,power_law_1.01,36.870502929687504
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,1,power_law_1.2,4.553241577148437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,1,power_law_1.2,5.794154052734375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,1,power_law_1.2,8.261052856445312
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,1,power_law_1.2,0.3539110565185547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,1,power_law_1.2,0.6609638214111329
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,1,power_law_1.2,0.24009216308593748
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,1,power_law_1.2,1.7116908264160156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,1,power_law_1.2,3.0448184204101563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,1,power_law_1.2,3.0642041015625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,1,power_law_1.2,3.1567449951171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,1,power_law_1.2,10.761788330078124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,1,power_law_1.2,0.9275769805908203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,1,power_law_1.2,3.281525268554687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,1,power_law_1.2,3.2321600341796874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,1,power_law_1.2,3.4890905761718747
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,1,power_law_1.2,3.52406982421875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,1,power_law_1.2,3.44716796875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,1,power_law_1.2,3.8163986206054688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,1,power_law_1.2,3.9573873901367187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,1,power_law_1.2,3.9341189575195314
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,1,power_law_1.2,4.448055725097656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,1,power_law_1.2,5.053048400878906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,1,power_law_1.2,5.199321899414063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,1,power_law_1.2,5.77216064453125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,1,power_law_1.2,7.170858764648438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,1,power_law_1.2,8.373125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,1,power_law_1.2,9.565504150390625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,16,1,power_law_1.2,13.338048095703124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,1,power_law_1.2,14.052337646484375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,1,power_law_1.2,16.587080078125002
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,1,power_law_1.01,0.0328985595703125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,1,power_law_1.01,0.03903680086135865
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,1,power_law_1.01,0.04398784160614014
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,1,power_law_1.01,0.040702080726623534
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,1,power_law_1.01,0.07456255912780761
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,1,power_law_1.01,0.07568960189819336
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,1,power_law_1.01,0.07609792232513428
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,1,power_law_1.01,0.07577343940734863
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,1,power_law_1.01,0.07505663871765136
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,1,power_law_1.01,0.0777510404586792
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,1,power_law_1.01,0.08020992279052734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,1,power_law_1.01,0.08239744186401368
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,1,power_law_1.01,0.08670656204223633
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,1,power_law_1.01,0.08822463989257813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,1,power_law_1.01,0.09433088302612305
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,1,power_law_1.01,0.09766400337219239
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,1,power_law_1.01,0.09857600212097169
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,1,power_law_1.01,0.11040127754211426
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,1,power_law_1.01,0.11956671714782714
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,1,power_law_1.01,0.14272768020629883
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,1,power_law_1.01,0.1716044807434082
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,1,power_law_1.01,0.24250560760498047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,1,power_law_1.2,22.31466552734375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,1,power_law_1.01,0.29651391983032227
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,1,power_law_1.01,0.41821952819824215
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,1,power_law_1.01,0.5370425415039063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,1,power_law_1.01,0.7688735961914063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,1,power_law_1.2,34.501445312499996
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,1,power_law_1.01,0.03331648111343384
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,1,power_law_1.01,0.03578111886978149
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,1,power_law_1.01,0.03616767883300781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,1,power_law_1.01,0.036595840454101566
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,1,power_law_1.01,0.042707200050354006
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,1,power_law_1.01,0.043553919792175294
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,1,power_law_1.01,0.04385727882385254
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,1,power_law_1.01,0.044492158889770504
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,1,power_law_1.01,0.04509759902954101
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,1,power_law_1.01,0.04558720111846924
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,1,power_law_1.01,0.04616127967834473
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,1,power_law_1.01,0.04713600158691406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,1,power_law_1.01,0.048519678115844726
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,1,power_law_1.01,1.0032972717285156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,1,power_law_1.01,0.053741440773010254
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,1,power_law_1.01,0.05664447784423828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,1,power_law_1.01,0.05827072143554688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,1,power_law_1.01,0.05593279838562012
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,1,power_law_1.01,0.06162303924560547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,1,power_law_1.01,0.07210559844970703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,1,power_law_1.01,0.08607487678527832
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,1,power_law_1.01,0.09904000282287598
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,1,power_law_1.01,0.14150400161743165
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,1,power_law_1.01,0.17174400329589842
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,1,power_law_1.01,0.24942911148071287
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,1,power_law_1.01,0.32308799743652344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,16,1,power_law_1.2,38.3763818359375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,32,1,power_law_1.01,1.2332064056396486
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,1,power_law_1.01,0.47150592803955077
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,1,power_law_1.01,0.6036102294921875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,32,1,power_law_1.01,0.0181548798084259
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,32,1,power_law_1.01,0.020638720989227293
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,32,1,power_law_1.01,0.020633599758148193
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,32,1,power_law_1.01,0.02060863971710205
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,32,1,power_law_1.01,0.020599679946899412
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,32,1,power_law_1.01,0.020636160373687745
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,32,1,power_law_1.01,0.02064255952835083
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,32,1,power_law_1.01,0.02097856044769287
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,32,1,power_law_1.01,0.021014399528503418
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,32,1,power_law_1.01,0.018989440202713013
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,32,1,power_law_1.01,0.7526879882812499
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,32,1,power_law_1.01,0.019839999675750734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,32,1,power_law_1.01,0.021891839504241943
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,32,1,power_law_1.01,0.02140928030014038
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,32,1,power_law_1.01,0.023165440559387206
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,32,1,power_law_1.01,0.023939199447631836
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,32,1,power_law_1.01,0.024830079078674315
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,32,1,power_law_1.01,0.02686208009719849
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,32,1,power_law_1.01,0.03299455881118775
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,32,1,power_law_1.01,0.03941888093948364
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,32,1,power_law_1.01,0.05032832145690917
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,32,1,power_law_1.01,0.05955455780029297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,32,1,power_law_1.01,0.08075776100158691
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,32,1,power_law_1.01,0.10050111770629884
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,32,1,power_law_1.01,0.13995712280273437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,32,1,power_law_1.01,0.1814406394958496
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,1,power_law_1.2,0.02757055997848511
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,1,power_law_1.2,0.037429759502410884
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,1,power_law_1.2,0.043381118774414064
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,1,power_law_1.2,0.0438374376296997
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,1,power_law_1.2,0.0616819190979004
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,1,power_law_1.2,0.06318016052246093
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,1,power_law_1.2,0.06425087928771973
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,1,power_law_1.2,0.06489088058471679
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,1,power_law_1.2,0.06684864044189454
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,1,power_law_1.2,0.07373439788818359
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,1,power_law_1.2,0.08055487632751465
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,1,power_law_1.2,0.08198335647583008
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,1,power_law_1.2,0.08469183921813965
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,1,power_law_1.2,0.08930432319641113
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,1,power_law_1.2,0.09114944458007812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,1,power_law_1.2,0.09673919677734374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,1,power_law_1.2,0.09845824241638183
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,32,1,power_law_1.01,0.2667321586608887
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,1,power_law_1.2,0.10872639656066893
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,1,power_law_1.2,0.12005632400512696
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,1,power_law_1.2,0.14289279937744143
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,1,power_law_1.2,0.1750271987915039
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,1,power_law_1.2,0.24163391113281252
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,32,1,power_law_1.01,0.34936641693115233
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,1,power_law_1.2,0.29760576248168946
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,1,power_law_1.2,0.41813632965087894
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,32,1,power_law_1.01,0.43766849517822265
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,1,power_law_1.2,0.5386009597778321
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,1,power_law_1.2,0.03334399938583374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,1,power_law_1.2,0.0341811203956604
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,1,power_law_1.2,0.03539583921432495
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,1,power_law_1.2,0.03625024080276489
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,1,power_law_1.2,0.04195136070251465
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,1,power_law_1.2,0.042789759635925295
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,1,power_law_1.2,0.04310592174530029
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,1,power_law_1.2,0.044430079460144045
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,1,power_law_1.2,0.04493440151214599
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,1,power_law_1.2,0.04537087917327881
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,1,power_law_1.2,0.046071681976318356
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,1,power_law_1.2,0.04725823879241943
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,1,power_law_1.2,0.04853951930999756
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,1,power_law_1.2,0.05332223892211914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,1,power_law_1.2,0.05705023765563964
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,1,power_law_1.2,0.059144320487976076
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,1,power_law_1.2,0.05620927810668945
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,1,power_law_1.2,0.061560959815979
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,1,power_law_1.2,0.07330495834350585
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,1,power_law_1.2,0.08502528190612793
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,1,power_law_1.2,0.7757785797119141
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,1,power_law_1.2,0.09932352066040039
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,1,power_law_1.2,0.14230079650878907
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,1,power_law_1.2,0.1714067268371582
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,1,power_law_1.2,0.2484614372253418
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,1,power_law_1.2,0.9983360290527343
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,1,power_law_1.2,0.3231398391723633
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,32,1,power_law_1.2,1.243367691040039
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,32,1,power_law_1.2,0.018895360231399535
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,32,1,power_law_1.2,0.01900928020477295
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,1,power_law_1.2,0.4714476776123047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,32,1,power_law_1.2,0.019351040124893186
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,32,1,power_law_1.2,0.02058624029159546
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,32,1,power_law_1.2,0.020606720447540285
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,32,1,power_law_1.2,0.02065471887588501
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,32,1,power_law_1.2,0.02064512014389038
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,32,1,power_law_1.2,0.020603520870208742
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,32,1,power_law_1.2,0.021012480258941653
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,32,1,power_law_1.2,0.02100543975830078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,32,1,power_law_1.2,0.021608960628509522
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,32,1,power_law_1.2,0.021841280460357666
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,32,1,power_law_1.2,0.02310080051422119
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,32,1,power_law_1.2,0.023921918869018555
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,32,1,power_law_1.2,0.025450239181518553
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,32,1,power_law_1.2,0.020227200984954834
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,32,1,power_law_1.2,0.033973119258880614
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,32,1,power_law_1.2,0.02724800109863281
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,1,power_law_1.2,0.6015705490112305
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,32,1,power_law_1.2,0.050369281768798825
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,32,1,power_law_1.2,0.060538239479064934
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,32,1,power_law_1.2,0.03943936109542846
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,32,1,power_law_1.2,0.7474508666992188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,32,1,power_law_1.2,0.08351679801940917
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,32,1,power_law_1.2,0.10402303695678712
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,32,1,power_law_1.2,0.14538496017456054
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,32,1,power_law_1.2,0.18832384109497072
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,1,power_law_1.01,0.05361087799072266
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,1,power_law_1.01,0.06154623985290527
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,1,power_law_1.01,0.0638483190536499
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,1,power_law_1.01,0.07150976181030273
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,1,power_law_1.01,0.11362751960754394
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,1,power_law_1.01,0.11489088058471679
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,1,power_law_1.01,0.11578495979309082
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,1,power_law_1.01,0.11651904106140136
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,1,power_law_1.01,0.11720959663391113
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,1,power_law_1.01,0.11824255943298341
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,1,power_law_1.01,0.12218751907348632
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,1,power_law_1.01,0.12484031677246095
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,1,power_law_1.01,0.1283353614807129
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,1,power_law_1.01,0.1312492847442627
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,1,power_law_1.01,0.13443776130676272
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,1,power_law_1.01,0.13892160415649416
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,1,power_law_1.01,0.14206591606140137
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,1,power_law_1.01,0.15665216445922853
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,1,power_law_1.01,0.17206016540527344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,1,power_law_1.01,0.20682304382324218
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,1,power_law_1.01,0.27224063873291016
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,1,power_law_1.01,0.3713420867919922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,32,1,power_law_1.2,0.27745216369628906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,1,power_law_1.01,0.46707008361816404
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,1,power_law_1.01,0.6591283416748046
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,1,power_law_1.01,0.8388044738769531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,32,1,power_law_1.2,0.36343360900878907
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,1,power_law_1.01,1.2203282928466797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,1,power_law_1.01,0.0370412802696228
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,1,power_law_1.01,0.040675201416015626
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,1,power_law_1.01,0.04192831993103027
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,1,power_law_1.01,0.04447296142578125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,1,power_law_1.01,0.05215616226196289
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,1,power_law_1.01,0.053034238815307624
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,1,power_law_1.01,0.05388160228729248
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,1,power_law_1.01,0.05426688194274902
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,1,power_law_1.01,0.05499392032623292
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,1,power_law_1.01,0.05534272193908692
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,1,power_law_1.01,0.05679872035980225
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,1,power_law_1.01,0.05852608203887939
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,1,power_law_1.01,0.061178879737854006
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,1,power_law_1.01,0.0628326416015625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,1,power_law_1.01,0.06641791820526123
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,1,power_law_1.01,0.06983424186706542
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,1,power_law_1.01,0.07307839870452881
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,1,power_law_1.01,0.08366335868835449
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,1,power_law_1.01,0.09561152458190918
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,1,power_law_1.01,0.11467071533203126
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,1,power_law_1.01,0.13324159622192383
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,1,power_law_1.01,1.5762757873535156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,1,power_law_1.01,0.1953785514831543
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,1,power_law_1.01,0.2369766426086426
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,32,1,power_law_1.2,0.4516089630126953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,1,power_law_1.01,0.3411052703857422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,1,power_law_1.01,0.4431110382080078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,32,1,power_law_1.01,1.948230438232422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,1,power_law_1.01,0.6473849487304688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,1,power_law_1.01,0.8309337615966796
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,32,1,power_law_1.01,0.019382400512695314
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,32,1,power_law_1.01,0.022756481170654298
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,32,1,power_law_1.01,0.023070719242095947
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,32,1,power_law_1.01,0.022760961055755615
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,32,1,power_law_1.01,0.02307904005050659
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,32,1,power_law_1.01,0.02278464078903198
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,32,1,power_law_1.01,0.02312448024749756
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,32,1,power_law_1.01,0.023208320140838623
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,32,1,power_law_1.01,0.023879680633544922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,32,1,power_law_1.01,0.020655360221862793
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,32,1,power_law_1.01,1.0279417419433594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,32,1,power_law_1.01,0.0210099196434021
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,32,1,power_law_1.01,0.026003839969635012
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,32,1,power_law_1.01,0.024384639263153075
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,32,1,power_law_1.01,0.027808001041412355
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,32,1,power_law_1.01,0.02988287925720215
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,32,1,power_law_1.01,0.030970880985260008
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,32,1,power_law_1.01,0.042872958183288574
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,32,1,power_law_1.01,0.035006079673767086
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.01,0.054551677703857424
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.01,0.0671776008605957
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.01,0.07892543792724609
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.01,0.10731776237487793
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.01,0.13787520408630372
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.01,0.19691328048706053
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.01,0.25424383163452147
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,1,power_law_1.2,0.055022082328796386
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,1,power_law_1.2,0.061700482368469235
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,1,power_law_1.2,0.062362880706787104
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,1,power_law_1.2,0.071495041847229
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,1,power_law_1.2,0.11344063758850098
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,1,power_law_1.2,0.11463680267333984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,1,power_law_1.2,0.11587008476257324
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,1,power_law_1.2,0.11643584251403809
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,1,power_law_1.2,0.11786687850952149
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,1,power_law_1.2,0.11881600379943849
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,1,power_law_1.2,0.12279616355895998
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,1,power_law_1.2,0.12581376075744627
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,1,power_law_1.2,0.128154878616333
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,1,power_law_1.2,0.1314367961883545
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,1,power_law_1.2,0.1348787212371826
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,1,power_law_1.2,0.1401203155517578
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,1,power_law_1.2,0.1422873592376709
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,1,power_law_1.2,0.15699456214904786
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,1,power_law_1.2,0.17293567657470704
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,1,power_law_1.2,0.20815744400024414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,1,power_law_1.2,0.2742393684387207
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.01,0.37286144256591797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,1,power_law_1.2,0.3725158309936524
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,1,power_law_1.2,0.4663065719604492
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.01,0.4952870559692383
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,1,power_law_1.2,0.6548384094238282
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,32,1,power_law_1.01,0.6176774215698242
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,1,power_law_1.2,0.837939224243164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,1,power_law_1.2,0.035774719715118405
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,1,power_law_1.2,0.039512319564819334
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,1,power_law_1.2,0.03984639883041382
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,1,power_law_1.2,0.04315904140472412
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,1,power_law_1.2,0.05172927856445313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,1,power_law_1.2,0.05251391887664795
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,1,power_law_1.2,0.05304128170013428
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,1,power_law_1.2,0.05438528060913086
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,1,power_law_1.2,0.05446080207824707
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,1,power_law_1.2,0.054640641212463384
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,1,power_law_1.2,0.055579519271850585
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,1,power_law_1.2,0.05731200218200684
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,1,power_law_1.2,0.0598419189453125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,1,power_law_1.2,0.06294976234436035
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,1,power_law_1.2,0.06730944156646729
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,1,power_law_1.2,0.07064064025878906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,1,power_law_1.2,0.07322175979614258
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,1,power_law_1.2,0.08334400177001952
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,1,power_law_1.2,0.0946662425994873
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,1,power_law_1.2,0.11308799743652345
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,1,power_law_1.2,1.2210118103027345
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,1,power_law_1.2,0.1326956844329834
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,1,power_law_1.2,0.19564096450805663
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,1,power_law_1.2,0.23603391647338867
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,1,power_law_1.2,1.5725453186035157
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,1,power_law_1.2,0.34136894226074216
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,1,power_law_1.2,0.4407724761962891
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,32,1,power_law_1.2,1.9613209533691407
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,32,1,power_law_1.2,0.0194214403629303
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,32,1,power_law_1.2,0.02067647933959961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,32,1,power_law_1.2,0.021033599376678466
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,32,1,power_law_1.2,0.022528638839721678
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,32,1,power_law_1.2,0.02266752004623413
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,32,1,power_law_1.2,0.022689919471740722
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,32,1,power_law_1.2,0.023045759201049804
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,32,1,power_law_1.2,0.022865920066833495
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,32,1,power_law_1.2,0.023114240169525145
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,32,1,power_law_1.2,0.023449599742889404
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,32,1,power_law_1.2,0.0238374400138855
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,32,1,power_law_1.2,0.02470720052719116
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,32,1,power_law_1.2,0.026325120925903323
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,32,1,power_law_1.2,0.027936639785766604
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,32,1,power_law_1.2,0.030028159618377688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,32,1,power_law_1.2,0.031690239906311035
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,1,power_law_1.2,0.6431865692138672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,32,1,power_law_1.2,0.043532800674438474
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,32,1,power_law_1.2,0.035190401077270506
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,1,power_law_1.2,0.831267852783203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.2,0.06895743846893311
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.2,0.05577023983001709
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.2,0.08198528289794922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,32,1,power_law_1.2,1.0282514953613282
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.2,0.11150143623352052
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.2,0.14158464431762696
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.2,0.20602176666259764
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.2,0.26333568572998045
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,1,power_law_1.01,0.0415180778503418
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,1,power_law_1.01,0.04598207950592041
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,1,power_law_1.01,0.07819392204284667
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,1,power_law_1.01,0.0970963191986084
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,1,power_law_1.01,0.12176192283630372
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,1,power_law_1.01,0.15294464111328127
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,1,power_law_1.01,0.16963136672973633
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,1,power_law_1.01,0.17745664596557617
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,1,power_law_1.01,0.1817888069152832
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,1,power_law_1.01,0.1852761650085449
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,1,power_law_1.01,0.1920518493652344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,1,power_law_1.01,0.20551551818847655
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,1,power_law_1.01,0.2136435127258301
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,1,power_law_1.01,0.23715263366699219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,1,power_law_1.01,0.3112716865539551
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.2,0.38874881744384765
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,1,power_law_1.01,0.32007808685302735
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,1,power_law_1.01,0.33438209533691404
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,1,power_law_1.01,0.3622585678100586
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,1,power_law_1.01,0.3902431869506836
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,1,power_law_1.01,0.4457478332519531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,1,power_law_1.01,0.5036569595336914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,1,power_law_1.01,0.6585081481933595
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,1,power_law_1.01,0.8189363098144531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,1,power_law_1.01,1.0692332458496092
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,1,power_law_1.01,1.3555410766601563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,1,power_law_1.01,1.889129638671875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,1,power_law_1.01,2.4539340209960936
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,1,power_law_1.01,0.0459987211227417
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,1,power_law_1.01,0.05214591979980469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,1,power_law_1.01,0.06849279880523682
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,1,power_law_1.01,0.0979366397857666
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,32,1,power_law_1.01,2.9994586181640623
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,1,power_law_1.01,0.1361753559112549
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,1,power_law_1.01,0.1892108726501465
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,1,power_law_1.01,0.21659711837768553
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,1,power_law_1.01,0.22806079864501952
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,1,power_law_1.01,0.2242207908630371
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,1,power_law_1.01,0.23119039535522462
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,1,power_law_1.01,0.23879039764404295
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,1,power_law_1.01,0.2476166343688965
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,1,power_law_1.01,0.2510111999511719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,1,power_law_1.01,0.2578233528137207
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,1,power_law_1.01,0.308078727722168
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,1,power_law_1.01,0.3146291160583496
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,1,power_law_1.01,0.3242995071411133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,1,power_law_1.01,0.3474905776977539
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,1,power_law_1.01,0.3733542251586914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,1,power_law_1.01,0.42502273559570314
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,1,power_law_1.01,0.4900249481201172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,1,power_law_1.01,0.6014035034179688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,1,power_law_1.01,0.7006515502929688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,1,power_law_1.01,0.9261759948730468
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,1,power_law_1.01,1.1581075286865234
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,1,power_law_1.01,1.6183526611328127
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,1,power_law_1.01,2.0429011535644532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,32,1,power_law_1.01,2.5035411071777345
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.2,0.5147910308837891
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,32,1,power_law_1.01,0.0771999979019165
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,32,1,power_law_1.01,0.10052224159240723
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,32,1,power_law_1.01,0.08719807624816894
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,32,1,power_law_1.01,0.07444352149963379
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,32,1,power_law_1.01,0.11643008232116699
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,32,1,power_law_1.01,0.13104576110839844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,32,1,power_law_1.01,0.16205375671386718
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,32,1,power_law_1.01,0.16809215545654296
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,32,1,power_law_1.01,0.17013248443603515
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,32,1,power_law_1.01,0.17334335327148437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,32,1,power_law_1.01,0.18193727493286133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,32,1,power_law_1.01,0.18756864547729493
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,32,1,power_law_1.01,0.19251455307006834
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,32,1,power_law_1.01,0.2010483169555664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,32,1,power_law_1.01,0.21133312225341797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,32,1,power_law_1.01,0.22060543060302734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,32,1,power_law_1.01,0.23715135574340823
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,32,1,power_law_1.2,0.6401708984375001
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,32,1,power_law_1.01,0.27933631896972655
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,32,1,power_law_1.01,0.4101523208618164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,32,1,power_law_1.01,0.3250374221801758
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,32,1,power_law_1.01,0.5047942352294922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,32,1,power_law_1.01,0.696622085571289
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,32,1,power_law_1.01,0.8987731170654296
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,32,1,power_law_1.01,1.261088638305664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,32,1,power_law_1.01,1.7678585815429688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,1,power_law_1.2,0.04680319786071777
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,1,power_law_1.2,0.04808832168579101
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,1,power_law_1.2,0.07845632076263429
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,1,power_law_1.2,0.09361087799072265
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,1,power_law_1.2,0.1148038387298584
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,1,power_law_1.2,0.1432691192626953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,1,power_law_1.2,0.16931072235107422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,1,power_law_1.2,0.17561216354370118
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,1,power_law_1.2,0.1802387237548828
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,1,power_law_1.2,0.186313591003418
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,1,power_law_1.2,0.19444543838500977
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,1,power_law_1.2,0.20658687591552732
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,1,power_law_1.2,0.21641471862792966
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,32,1,power_law_1.01,3.4947225952148435
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,1,power_law_1.2,0.24338239669799805
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,1,power_law_1.2,0.3143692779541015
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,1,power_law_1.2,0.3364147186279297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,1,power_law_1.2,0.32268352508544923
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,1,power_law_1.2,0.36686977386474606
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,1,power_law_1.2,0.3930815887451172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,1,power_law_1.2,0.45250495910644534
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,32,1,power_law_1.01,2.6394772338867187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,1,power_law_1.2,0.5060736083984375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,1,power_law_1.2,0.6646195220947265
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,1,power_law_1.2,0.8195053100585937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,1,power_law_1.2,1.080145263671875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,1,power_law_1.2,1.367263946533203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,1,power_law_1.2,1.9000460815429687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,1,power_law_1.2,2.4646067810058594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,1,power_law_1.2,0.04561920166015625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,1,power_law_1.2,0.0517958402633667
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,1,power_law_1.2,0.09339712142944337
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,1,power_law_1.2,0.12746432304382324
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,1,power_law_1.2,0.07128064155578613
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,7168,2048,8,256,32,1,power_law_1.01,4.136628417968749
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,32,1,power_law_1.2,3.0382879638671874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,1,power_law_1.2,0.1783897590637207
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,1,power_law_1.2,0.21395135879516602
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,1,power_law_1.2,0.22245504379272463
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,1,power_law_1.2,0.22708480834960937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,1,power_law_1.2,0.23716608047485352
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,1,power_law_1.2,0.23193279266357422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,1,power_law_1.2,0.24373119354248046
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,1,power_law_1.2,0.2521452713012695
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,1,power_law_1.2,0.26048959732055665
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,1,power_law_1.2,0.31215232849121094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,1,power_law_1.2,0.31855871200561525
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,1,power_law_1.2,0.33202560424804684
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,1,power_law_1.2,0.35955329895019533
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,1,power_law_1.2,0.38645694732666014
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,1,power_law_1.2,0.4423775863647461
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,1,power_law_1.2,0.5106745529174804
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,1,power_law_1.2,0.6149843215942383
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,1,power_law_1.2,0.7421625518798829
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,1,power_law_1.2,0.9347090911865233
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,1,power_law_1.2,1.1793939208984374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,1,power_law_1.2,1.6255839538574217
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,32,1,power_law_1.2,0.07463615894317627
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,1,power_law_1.2,2.086827545166016
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,32,1,power_law_1.2,0.07751423835754394
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,32,1,power_law_1.2,0.08652735710144042
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,32,1,power_law_1.2,2.504997100830078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,32,1,power_law_1.2,0.11208640098571779
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,32,1,power_law_1.2,0.12774271965026857
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,32,1,power_law_1.2,0.1604844856262207
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,32,1,power_law_1.2,0.16681600570678712
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,32,1,power_law_1.2,0.17164672851562499
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,32,1,power_law_1.2,0.17496255874633787
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,32,1,power_law_1.2,0.1823129653930664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,32,1,power_law_1.2,0.19001152038574218
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,32,1,power_law_1.2,0.19640512466430665
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,32,1,power_law_1.2,0.21030336380004883
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,32,1,power_law_1.2,0.2240671920776367
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,32,1,power_law_1.2,0.23531967163085937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,32,1,power_law_1.2,0.26681856155395506
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,32,1,power_law_1.2,0.09857215881347656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,32,1,power_law_1.2,0.3303865432739258
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,32,1,power_law_1.2,0.38400257110595704
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,32,1,power_law_1.2,0.5208678436279297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,32,1,power_law_1.2,0.6809977722167969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,32,1,power_law_1.2,0.9844108581542969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,32,1,power_law_1.2,1.3062860107421874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,32,1,power_law_1.2,1.9810214233398438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,1,power_law_1.01,0.03556479930877686
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,1,power_law_1.01,0.05352511882781983
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,1,power_law_1.01,0.05976895809173584
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,1,power_law_1.01,0.06831232070922852
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,1,power_law_1.01,0.08270976066589356
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,1,power_law_1.01,0.0813036823272705
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,1,power_law_1.01,0.08379839897155762
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,1,power_law_1.01,0.08646528244018555
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,1,power_law_1.01,0.08862015724182129
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,1,power_law_1.01,0.09343808174133302
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,1,power_law_1.01,0.1318879985809326
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,1,power_law_1.01,0.13444416046142577
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,1,power_law_1.01,0.14004159927368165
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,1,power_law_1.01,0.1431545639038086
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,32,1,power_law_1.2,2.7397393798828125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,1,power_law_1.01,0.1470252799987793
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,1,power_law_1.01,0.1741971206665039
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,1,power_law_1.01,0.1591763210296631
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,1,power_law_1.01,0.18978048324584962
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,1,power_law_1.01,0.22477888107299804
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,1,power_law_1.01,0.25989376068115233
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,1,power_law_1.01,0.3392505645751953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,1,power_law_1.01,0.42946624755859375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,1,power_law_1.01,0.5915590286254883
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,1,power_law_1.01,0.7537760162353516
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,1,power_law_1.01,1.086407699584961
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,1,power_law_1.01,1.4060762023925781
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,1,power_law_1.01,0.036208639144897456
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,1,power_law_1.01,0.040629119873046876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,1,power_law_1.01,0.04884352207183838
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,1,power_law_1.01,0.05714303970336913
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,32,1,power_law_1.01,1.7418432617187498
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,1,power_law_1.01,0.07140416145324707
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,1,power_law_1.01,0.08257535934448243
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,1,power_law_1.01,0.08370368003845215
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,1,power_law_1.01,0.08542464256286622
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,1,power_law_1.01,0.08674688339233398
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,1,power_law_1.01,0.08837759971618653
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,1,power_law_1.01,0.09142463684082032
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,1,power_law_1.01,0.1081824016571045
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,1,power_law_1.01,0.1117587184906006
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,1,power_law_1.01,0.11723903656005859
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,1,power_law_1.01,0.1189964771270752
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,1,power_law_1.01,0.12280063629150391
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,1,power_law_1.01,0.1332748794555664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,1,power_law_1.01,0.14709376335144042
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,1,power_law_1.01,0.17243904113769531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,1,power_law_1.01,0.20651199340820314
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,1,power_law_1.01,0.24172096252441405
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,1,power_law_1.01,0.30732160568237304
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,1,power_law_1.01,0.3821664047241211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,1,power_law_1.01,0.5286483383178711
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,1,power_law_1.01,0.6549324798583984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,1,power_law_1.01,0.9408787536621095
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,1,power_law_1.01,1.2183763122558593
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,32,1,power_law_1.01,1.4970367431640625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,1,power_law_1.2,0.031379199028015135
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,1,power_law_1.2,0.03820352077484131
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,1,power_law_1.2,0.05305088043212891
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,1,power_law_1.2,0.054250240325927734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,1,power_law_1.2,0.06553791999816895
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,1,power_law_1.2,0.08074239730834962
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,1,power_law_1.2,0.08264512062072754
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,1,power_law_1.2,0.08498432159423827
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,1,power_law_1.2,0.08864192008972169
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,1,power_law_1.2,0.08997440338134766
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,1,power_law_1.2,0.09441023826599121
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,1,power_law_1.2,0.13309120178222655
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,1,power_law_1.2,0.135413761138916
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,1,power_law_1.2,0.14131584167480468
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,1,power_law_1.2,0.14355199813842773
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,1,power_law_1.2,0.1481766414642334
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,1,power_law_1.2,0.17533248901367188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,1,power_law_1.2,0.15632448196411133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,1,power_law_1.2,0.19262399673461914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,1,power_law_1.2,0.22632511138916014
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,1,power_law_1.2,0.2608595275878906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,1,power_law_1.2,0.342492790222168
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,1,power_law_1.2,0.42954368591308595
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,1,power_law_1.2,0.5895955276489258
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,1,power_law_1.2,0.7579571533203124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,1,power_law_1.2,1.0887769317626952
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,1,power_law_1.2,0.03663232088088989
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,1,power_law_1.2,0.041091837882995606
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,1,power_law_1.2,0.04772799968719482
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,1,power_law_1.2,0.05542079925537109
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,1,power_law_1.2,0.06844543933868408
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,1,power_law_1.2,0.08376895904541015
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,1,power_law_1.2,0.08432767868041993
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,1,power_law_1.2,0.08582271575927734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,1,power_law_1.2,0.08868032455444336
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,1,power_law_1.2,0.08790016174316406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,1,power_law_1.2,0.0907692813873291
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,1,power_law_1.2,0.10989439964294434
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,1,power_law_1.2,0.11051456451416015
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,1,power_law_1.2,0.11843775749206542
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,1,power_law_1.2,0.12273344039916992
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,1,power_law_1.2,0.12558336257934571
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,1,power_law_1.2,0.13538880348205568
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,1,power_law_1.2,0.17878976821899414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,1,power_law_1.2,0.21258432388305665
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,1,power_law_1.2,0.24491647720336912
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,1,power_law_1.2,0.3083872032165528
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,1,power_law_1.2,0.1473094367980957
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,1,power_law_1.2,0.38800064086914066
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,1,power_law_1.2,0.522630386352539
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,1,power_law_1.2,0.6566175842285157
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,1,power_law_1.2,1.4113145446777344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,1,power_law_1.2,0.9310393524169921
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,32,1,power_law_1.2,1.7476626586914061
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,1,power_law_1.2,1.2198233795166016
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,32,1,power_law_1.2,1.4916288757324219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,1,power_law_1.01,0.045416960716247554
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,1,power_law_1.01,0.05385791778564453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,1,power_law_1.01,0.07671999931335449
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,1,power_law_1.01,0.09023360252380372
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,1,power_law_1.01,0.10860223770141601
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,1,power_law_1.01,0.13515392303466797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,1,power_law_1.01,0.1402726364135742
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,1,power_law_1.01,0.14545023918151856
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,1,power_law_1.01,0.14989184379577636
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,1,power_law_1.01,0.15369600296020508
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,1,power_law_1.01,0.1650771141052246
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,1,power_law_1.01,0.17590335845947264
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,1,power_law_1.01,0.22148672103881836
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,1,power_law_1.01,0.22869184494018552
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,1,power_law_1.01,0.23249727249145508
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,1,power_law_1.01,0.2379360008239746
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,1,power_law_1.01,0.253375358581543
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,1,power_law_1.01,0.2807366371154785
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,1,power_law_1.01,0.30842432022094723
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,1,power_law_1.01,0.36758529663085937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,1,power_law_1.01,0.45895935058593745
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,1,power_law_1.01,0.607133445739746
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,1,power_law_1.01,0.7378656005859374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,1,power_law_1.01,1.0183724975585937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,1,power_law_1.01,1.2980575561523438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,1,power_law_1.01,1.8498649597167969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,1,power_law_1.01,2.408307800292969
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,1,power_law_1.01,0.0463808012008667
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,32,1,power_law_1.01,2.969212951660156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,1,power_law_1.01,0.05134592056274414
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,1,power_law_1.01,0.06987071990966796
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,1,power_law_1.01,0.08817343711853028
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,1,power_law_1.01,0.11150336265563965
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,1,power_law_1.01,0.15896896362304686
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,1,power_law_1.01,0.15288640022277833
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,1,power_law_1.01,0.15665984153747559
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,1,power_law_1.01,0.16095935821533203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,1,power_law_1.01,0.16268287658691408
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,1,power_law_1.01,0.16739519119262697
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,1,power_law_1.01,0.1787609672546387
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,1,power_law_1.01,0.19826175689697265
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,1,power_law_1.01,0.21331008911132812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,1,power_law_1.01,0.21282623291015623
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,1,power_law_1.01,0.20414144515991212
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,1,power_law_1.01,0.23563711166381837
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,1,power_law_1.01,0.25910015106201173
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,1,power_law_1.01,0.2792742347717285
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,1,power_law_1.01,0.32640064239501954
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,1,power_law_1.01,0.3810060882568359
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,1,power_law_1.01,0.49230209350585935
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,1,power_law_1.01,0.5937401580810546
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,1,power_law_1.01,0.8164012908935547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,1,power_law_1.01,1.0222278594970704
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,1,power_law_1.01,1.434267578125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,1,power_law_1.01,1.85845947265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,32,1,power_law_1.01,2.291443176269531
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,1,power_law_1.2,0.04151360034942627
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,1,power_law_1.2,0.06881792068481445
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,1,power_law_1.2,0.07644288063049316
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,1,power_law_1.2,0.08865407943725587
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,1,power_law_1.2,0.10857024192810058
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,1,power_law_1.2,0.13961855888366698
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,1,power_law_1.2,0.13798336029052732
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,1,power_law_1.2,0.14521087646484374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,1,power_law_1.2,0.15049471855163574
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,1,power_law_1.2,0.15659968376159666
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,1,power_law_1.2,0.1665088081359863
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,1,power_law_1.2,0.1768320083618164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,1,power_law_1.2,0.222229118347168
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,1,power_law_1.2,0.2291270446777344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,1,power_law_1.2,0.23405952453613282
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,1,power_law_1.2,0.23969856262207032
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,1,power_law_1.2,0.2544825553894043
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,1,power_law_1.2,0.2805561637878418
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,1,power_law_1.2,0.30871871948242186
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,1,power_law_1.2,0.3775187301635742
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,1,power_law_1.2,0.4610329437255859
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,1,power_law_1.2,0.6068012619018555
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,1,power_law_1.2,0.7389164733886718
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,1,power_law_1.2,1.0199910736083984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,1,power_law_1.2,1.2942559814453125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,1,power_law_1.2,0.04562880039215088
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,1,power_law_1.2,0.0522054386138916
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,1,power_law_1.2,0.06873536109924316
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,1,power_law_1.2,0.08439104080200195
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,1,power_law_1.2,0.1122214412689209
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,1,power_law_1.2,0.15775296211242676
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,1,power_law_1.2,0.1512831974029541
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,1,power_law_1.2,0.15705663681030274
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,1,power_law_1.2,1.8577145385742189
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,1,power_law_1.2,0.16334463119506837
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,1,power_law_1.2,0.16504447937011718
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,1,power_law_1.2,0.17130367279052733
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,1,power_law_1.2,0.1773324775695801
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,1,power_law_1.2,0.2128371238708496
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,1,power_law_1.2,0.2142553520202637
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,1,power_law_1.2,0.22284543991088865
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,1,power_law_1.2,0.2307423973083496
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,1,power_law_1.2,0.24159616470336914
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,1,power_law_1.2,0.26137855529785153
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,1,power_law_1.2,0.282619514465332
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,1,power_law_1.2,0.3329593658447266
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,1,power_law_1.2,0.3821241760253906
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,1,power_law_1.2,0.5070982360839844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,1,power_law_1.2,0.5956812667846679
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,1,power_law_1.2,0.8203437042236328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,1,power_law_1.2,1.0205004882812498
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,1,power_law_1.2,2.404694366455078
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,1,power_law_1.2,1.4374560546875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,32,1,power_law_1.2,2.9759616088867187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,1,power_law_1.2,1.857093048095703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,32,1,power_law_1.2,2.2889932250976566
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,1,power_law_1.01,0.04888192176818847
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,1,power_law_1.01,0.050954880714416506
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,1,power_law_1.01,0.06849728107452394
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,1,power_law_1.01,0.09909055709838868
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,1,power_law_1.01,0.12928832054138184
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,1,power_law_1.01,0.1669363212585449
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,1,power_law_1.01,0.19456192016601564
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,1,power_law_1.01,0.22429824829101563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,1,power_law_1.01,0.23257024765014647
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,1,power_law_1.01,0.24313407897949219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,1,power_law_1.01,0.2506719970703125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,1,power_law_1.01,0.25630399703979495
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,1,power_law_1.01,0.2732358360290527
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,1,power_law_1.01,0.2994508743286133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,1,power_law_1.01,0.3176492881774903
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,1,power_law_1.01,0.3367206573486328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,1,power_law_1.01,0.4176505661010742
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,1,power_law_1.01,0.4781779098510742
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,1,power_law_1.01,0.4470816040039063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,1,power_law_1.01,0.5402867126464843
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,1,power_law_1.01,0.6121292877197265
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,1,power_law_1.01,0.8076518249511718
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,1,power_law_1.01,0.9238130950927734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,1,power_law_1.01,1.2085330963134766
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,1,power_law_1.01,1.4919949340820313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,1,power_law_1.01,2.073222351074219
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,1,power_law_1.01,2.640093994140625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,1,power_law_1.01,0.048888959884643556
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,1,power_law_1.01,0.05474175930023193
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,1,power_law_1.01,0.06654719829559326
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,1,power_law_1.01,0.10167231559753417
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,32,1,power_law_1.01,3.2311590576171874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,1,power_law_1.01,0.151276159286499
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,1,power_law_1.01,0.2098150444030762
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,1,power_law_1.01,0.2546553611755371
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,1,power_law_1.01,0.2932460784912109
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,1,power_law_1.01,0.30195775985717777
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,1,power_law_1.01,0.3196691131591797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,1,power_law_1.01,0.31138303756713864
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,1,power_law_1.01,0.3258086395263672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,1,power_law_1.01,0.3362752151489258
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,1,power_law_1.01,0.3551136016845703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,1,power_law_1.01,0.37052799224853517
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,1,power_law_1.01,0.4421203231811523
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,1,power_law_1.01,0.4662483215332031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,1,power_law_1.01,0.48919105529785156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,1,power_law_1.01,0.5442335891723633
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,1,power_law_1.01,0.601152000427246
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,1,power_law_1.01,0.7184710693359375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,1,power_law_1.01,0.8400121307373046
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,1,power_law_1.01,1.0890252685546875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,1,power_law_1.01,1.3141253662109373
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,1,power_law_1.01,1.7824748229980467
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,1,power_law_1.01,2.2638368225097656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,32,1,power_law_1.01,2.731412353515625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,32,1,power_law_1.01,0.09365376472473144
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,32,1,power_law_1.01,0.09327360153198241
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,32,1,power_law_1.01,0.09569984436035156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,32,1,power_law_1.01,0.09922176361083984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,32,1,power_law_1.01,0.09442432403564453
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,32,1,power_law_1.01,0.10585344314575196
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,32,1,power_law_1.01,0.11280384063720703
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,32,1,power_law_1.01,0.11935040473937988
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,32,1,power_law_1.01,0.12629183769226074
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,32,1,power_law_1.01,0.13289792060852051
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,32,1,power_law_1.01,0.1458463954925537
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,32,1,power_law_1.01,0.17201984405517579
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,32,1,power_law_1.01,0.1816511917114258
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,32,1,power_law_1.01,0.22050559997558591
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,32,1,power_law_1.01,0.288526725769043
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,32,1,power_law_1.01,0.3351155090332031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,32,1,power_law_1.01,0.43010112762451175
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,32,1,power_law_1.01,0.600640640258789
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,32,1,power_law_1.01,0.7675794982910157
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,32,1,power_law_1.01,1.1113235473632812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,32,1,power_law_1.01,1.4476039123535156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,32,1,power_law_1.01,2.142722625732422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,32,1,power_law_1.01,2.8181491088867188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,32,1,power_law_1.01,4.2340063476562495
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,1,power_law_1.2,0.047181439399719236
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,1,power_law_1.2,0.04850048065185547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,1,power_law_1.2,0.0998367977142334
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,1,power_law_1.2,0.08126144409179688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,1,power_law_1.2,0.1258803176879883
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,1,power_law_1.2,0.16066688537597656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,1,power_law_1.2,0.1868819236755371
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,1,power_law_1.2,0.23220415115356446
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,1,power_law_1.2,0.22229696273803712
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,1,power_law_1.2,0.2369913673400879
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,1,power_law_1.2,0.24383232116699222
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,1,power_law_1.2,0.26471424102783203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,1,power_law_1.2,0.27138944625854494
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,1,power_law_1.2,0.3032851219177246
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,1,power_law_1.2,0.3217107009887695
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,1,power_law_1.2,0.33774017333984374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,1,power_law_1.2,0.4214374542236328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,1,power_law_1.2,0.45129024505615234
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,1,power_law_1.2,0.48172798156738283
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,1,power_law_1.2,0.5511347198486328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,1,power_law_1.2,0.632734718322754
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,1,power_law_1.2,0.8130931091308593
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,1,power_law_1.2,0.9342272186279297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,1,power_law_1.2,1.2193145751953125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,1,power_law_1.2,1.5043814086914062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,1,power_law_1.2,2.0855398559570313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,1,power_law_1.2,2.665818786621094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,1,power_law_1.2,0.049683837890625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,1,power_law_1.2,0.05441728115081788
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,1,power_law_1.2,0.07078207969665527
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,1,power_law_1.2,0.10003840446472168
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,1,power_law_1.2,0.14273471832275392
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,32,1,power_law_1.2,3.259047546386719
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,1,power_law_1.2,0.20271488189697268
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,1,power_law_1.2,0.2461702346801758
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,1,power_law_1.2,0.2902892875671387
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,1,power_law_1.2,0.29921920776367184
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,1,power_law_1.2,0.3053254318237305
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,1,power_law_1.2,0.3140326309204101
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,1,power_law_1.2,0.32476158142089845
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,1,power_law_1.2,0.33509761810302735
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,1,power_law_1.2,0.35296447753906246
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,1,power_law_1.2,0.3578611373901367
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,1,power_law_1.2,0.3715372848510742
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,1,power_law_1.2,0.44275585174560544
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,1,power_law_1.2,0.47651840209960933
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,1,power_law_1.2,0.5030739212036133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,1,power_law_1.2,0.5628870391845704
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,1,power_law_1.2,0.6222508621215821
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,1,power_law_1.2,0.7607698822021485
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,1,power_law_1.2,0.8705670166015625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,1,power_law_1.2,1.113966064453125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,1,power_law_1.2,1.353317108154297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,1,power_law_1.2,1.8245823669433594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,1,power_law_1.2,2.2940960693359376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,32,1,power_law_1.2,2.7734100341796877
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,32,1,power_law_1.2,0.09385984420776368
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,32,1,power_law_1.2,0.0942796802520752
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,32,1,power_law_1.2,0.09598719596862793
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,32,1,power_law_1.2,0.09924415588378907
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,32,1,power_law_1.2,0.1061241626739502
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,32,1,power_law_1.2,0.11238016128540038
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,32,1,power_law_1.2,0.11969152450561524
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,32,1,power_law_1.2,0.12625087738037108
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,32,1,power_law_1.2,0.13278335571289063
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,32,1,power_law_1.2,0.14600576400756834
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,32,1,power_law_1.2,0.1881452751159668
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,32,1,power_law_1.2,0.09376704216003418
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,32,1,power_law_1.2,0.23484800338745115
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,32,1,power_law_1.2,0.2895609664916992
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,32,1,power_law_1.2,0.3368313598632812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,32,1,power_law_1.2,0.165599365234375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,32,1,power_law_1.2,0.4300902557373047
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,32,1,power_law_1.2,0.6005920028686523
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,32,1,power_law_1.2,0.7695481872558594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,32,1,power_law_1.2,1.1065888214111328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,32,1,power_law_1.2,1.4450028991699218
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,32,1,power_law_1.2,2.1443295288085937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,32,1,power_law_1.2,2.8274725341796874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,1,power_law_1.01,0.02883647918701172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,32,1,power_law_1.2,4.222255249023438
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,1,power_law_1.01,0.03624576091766357
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,1,power_law_1.01,0.05929215908050537
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,1,power_law_1.01,0.07100287914276124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,1,power_law_1.01,0.04887167930603027
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,1,power_law_1.01,0.0973356819152832
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,1,power_law_1.01,0.10105088233947754
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,1,power_law_1.01,0.0919052791595459
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,1,power_law_1.01,0.10529855728149413
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,1,power_law_1.01,0.10930751800537108
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,1,power_law_1.01,0.11640576362609863
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,1,power_law_1.01,0.13005120277404786
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,1,power_law_1.01,0.13496383666992187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,1,power_law_1.01,0.1368064022064209
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,1,power_law_1.01,0.14094783782958983
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,1,power_law_1.01,0.14924991607666016
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,1,power_law_1.01,0.12865663528442384
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,1,power_law_1.01,0.1658905601501465
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,1,power_law_1.01,0.18252031326293944
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,1,power_law_1.01,0.23473920822143554
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,1,power_law_1.01,0.27697727203369144
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,1,power_law_1.01,0.3377715301513672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,1,power_law_1.01,0.4268096160888672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,1,power_law_1.01,0.5590617752075195
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,1,power_law_1.01,0.7154950714111328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,1,power_law_1.01,1.029731216430664
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,1,power_law_1.01,1.3435314941406251
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,1,power_law_1.2,0.03079999923706055
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,1,power_law_1.2,0.03298687934875488
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,1,power_law_1.2,0.0409119987487793
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,32,1,power_law_1.01,1.6429965209960937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,1,power_law_1.2,0.05545599937438965
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,1,power_law_1.2,0.0906873607635498
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,1,power_law_1.2,0.09744576454162598
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,1,power_law_1.2,0.10322112083435057
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,1,power_law_1.2,0.10698816299438477
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,1,power_law_1.2,0.11201408386230469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,1,power_law_1.2,0.06690624237060547
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,1,power_law_1.2,0.11865023612976074
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,1,power_law_1.2,0.1303756809234619
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,1,power_law_1.2,0.13349632263183592
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,1,power_law_1.2,0.1384505558013916
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,1,power_law_1.2,0.13896384239196777
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,1,power_law_1.2,0.1445587158203125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,1,power_law_1.2,0.15156352043151855
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,1,power_law_1.2,0.16818687438964844
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,1,power_law_1.2,0.18381248474121095
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,1,power_law_1.2,0.2456268882751465
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,1,power_law_1.2,0.27822208404541016
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,1,power_law_1.2,0.34055553436279296
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,1,power_law_1.2,0.425214729309082
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,1,power_law_1.2,0.5600332641601563
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,1,power_law_1.2,0.7178387451171875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,1,power_law_1.2,1.032526092529297
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,1,power_law_1.2,1.354204864501953
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,1,power_law_1.01,0.059183359146118164
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,1,power_law_1.01,0.08928256034851074
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,1,power_law_1.01,0.32048702239990234
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,1,power_law_1.01,0.5073286437988281
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,1,power_law_1.01,0.5436492919921875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,1,power_law_1.01,0.56644287109375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,1,power_law_1.01,0.5886707305908203
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,1,power_law_1.01,0.6074764633178711
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,1,power_law_1.01,0.14788864135742188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,1,power_law_1.01,0.22489791870117187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,1,power_law_1.01,0.6508633422851562
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,32,1,power_law_1.2,1.6471270751953124
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,1,power_law_1.01,0.8636921691894532
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,1,power_law_1.01,0.7021107482910156
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,1,power_law_1.01,0.9268172454833984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,1,power_law_1.01,1.0299545288085938
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,1,power_law_1.01,0.749056625366211
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,1,power_law_1.01,1.2233984375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,1,power_law_1.01,0.9624684906005859
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,1,power_law_1.01,1.0573497772216798
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,1,power_law_1.01,1.55910400390625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,1,power_law_1.01,2.066100463867188
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,1,power_law_1.01,2.539087982177734
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,1,power_law_1.01,1.297615966796875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,1,power_law_1.01,3.555721740722656
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,1,power_law_1.01,4.5331423950195315
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,1,power_law_1.01,6.546027221679688
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,1,power_law_1.01,0.1794976043701172
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,1,power_law_1.01,8.575663452148437
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,1,power_law_1.01,0.5228102493286133
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,32,1,power_law_1.01,10.6433251953125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,1,power_law_1.01,1.118564453125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,1,power_law_1.01,1.9510188293457031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,1,power_law_1.01,2.419523162841797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,1,power_law_1.01,2.1698880004882812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,1,power_law_1.01,2.5818322753906253
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,1,power_law_1.01,2.245886077880859
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,1,power_law_1.01,0.23820991516113282
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,1,power_law_1.01,2.364017333984375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,1,power_law_1.01,2.7054693603515623
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,1,power_law_1.01,0.6930284881591797
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,1,power_law_1.01,2.467926330566406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,1,power_law_1.01,2.9328839111328127
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,1,power_law_1.01,2.8486700439453125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,1,power_law_1.01,3.2060076904296877
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,1,power_law_1.01,4.1378475952148435
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,1,power_law_1.01,4.036535034179687
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,1,power_law_1.01,1.9532447814941407
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,1,power_law_1.01,5.3219757080078125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,1,power_law_1.01,2.1902149963378905
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,1,power_law_1.01,6.202991333007812
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,1,power_law_1.01,10.848514404296875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,1,power_law_1.01,17.7012548828125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,1,power_law_1.01,6.166218872070313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,1,power_law_1.2,0.06154687881469727
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,1,power_law_1.2,0.09937600135803223
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,1,power_law_1.2,0.15040512084960939
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,1,power_law_1.2,0.21479360580444337
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,1,power_law_1.2,0.29202688217163086
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,1,power_law_1.01,24.85330322265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,1,power_law_1.2,0.4999308776855469
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,1,power_law_1.2,0.5464172744750977
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,1,power_law_1.01,12.2307861328125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,1,power_law_1.2,0.5612761688232422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,1,power_law_1.2,0.5890047836303711
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,1,power_law_1.2,0.6246156692504883
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,1,power_law_1.2,0.6664678192138671
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,1,power_law_1.2,0.7315948486328125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,1,power_law_1.2,0.7807334136962891
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,1,power_law_1.2,0.8881279754638672
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,1,power_law_1.2,0.9703001403808594
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,1,power_law_1.2,1.0624384307861328
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,1,power_law_1.2,1.2501773071289062
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,1,power_law_1.2,0.9904038238525391
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,1,power_law_1.2,1.0828729248046876
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,1,power_law_1.2,1.3463104248046875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,1,power_law_1.2,1.6038482666015625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,1,power_law_1.2,2.0869708251953125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,32,1,power_law_1.01,28.0950927734375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,1,power_law_1.2,2.5586566162109374
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,1,power_law_1.2,3.582673950195313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,1,power_law_1.2,4.563055419921875
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,1,power_law_1.2,6.595765991210937
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,1,power_law_1.2,0.27617023468017576
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,1,power_law_1.2,0.49537086486816406
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,1,power_law_1.2,0.8946688079833984
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,1,power_law_1.2,1.2223494720458983
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,1,power_law_1.2,1.617442626953125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,1,power_law_1.2,1.8001863098144533
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,1,power_law_1.2,0.14781567573547363
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,1,power_law_1.2,1.847546844482422
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,1,power_law_1.2,1.8466790771484376
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,1,power_law_1.2,1.9815980529785158
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,1,power_law_1.2,1.9562342834472655
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,1,power_law_1.2,2.046067199707031
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,1,power_law_1.2,2.033665924072266
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,1,power_law_1.2,2.187838134765625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,1,power_law_1.2,2.2038426208496094
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,1,power_law_1.2,8.628932495117187
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,1,power_law_1.2,2.320100555419922
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,1,power_law_1.2,2.461719665527344
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,1,power_law_1.2,3.103765869140625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,1,power_law_1.2,3.6158380126953125
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,1,power_law_1.2,3.746589965820313
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,1,power_law_1.2,4.325349121093749
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,1,power_law_1.2,6.088580322265625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,1,power_law_1.2,6.405139770507813
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,1,power_law_1.2,8.463971557617189
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,1,power_law_1.2,12.456920166015625
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,32,1,power_law_1.2,10.723162841796874
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,1,power_law_1.2,15.886087646484375
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,1,power_law_1.2,19.238581542968753
SGLang,0.5.9,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,32,1,power_law_1.2,28.313193359375
