Atlas

Benchmarks

← All benchmarks

KernelBench Mega — H100 Best Decode Speedup

Code · 2026-06-18

KernelBench Mega best valid decode speedup on the H100 PCIe board. The live headline metric uses the Kimi-Linear W4A16 decode problem and excludes the sparsely attempted RL Grid PPO cell from the performance aggregate.

Top models (higher is better)

ModelScore
Opus 4.815.5
Kimi K314.8
GPT-5.55.6
Composer 2.51.8
DeepSeek-V4-Pro1.4
Loading Atlas data…