List your company

End-to-End Inference Performance on LongBench single-node

0.79TTFT (Mean)FoE
0.67321.46162.253.0384May 7, 2026

Evaluation results

MethodLinks
FoEScale=1.5, Architecture=FoE2026.050.790.7441.6270.3340.3330.3658685.694.1
FoEScale=0.75, Architecture=FoE2026.051.0410.9792.1160.4580.4580.502117.9117.8129
FoEScale=0.5, Architecture=FoE2026.051.3941.2133.5280.5660.5960.639145.7153.8164.3
MoEScale=1.5, Architecture=MoE2026.052.2642.1185.1280.3710.360.55896.993.8146.7
MoEScale=0.75, Architecture=MoE2026.053.2333.1475.8560.8410.8691.033217.7225.2266.6
MoEScale=0.5, Architecture=MoE2026.053.713.6236.4370.780.7830.979202.7203.1253.3