KernelBench CUDA Normalized (B200)
Code · 2026-09-30
KernelBench CUDA B200 board: mean audited valid cell score relative to the board-best score per problem, scaled to 0–100 over four problems. Missing/invalid cells contribute zero. Best published cells may come from different harness routes and reasoning settings of the same model.
Top models (higher is better)
| Model | Score |
|---|---|
| Opus 4.8 | 71.3 |
| Claude Fable 5 | 56.4 |
| Grok 4.5 | 39.8 |
| Kimi K3 | 10.7 |