Kernel Arena - KernelBench HIP (source min run): leaderboard
Metric: Mean Correctness+Speedup. Source: github.com. Saturation forecast: Around May 2027. 4 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.5 | 2.88 |
| 2 | GPT-5.2 | 2.84 |
| 3 | GLM-4.7 | 2.39 |
| 4 | Grok 4.1 | 2.27 |
Interactive version: theaggregate.ai/benchmark?slug=kernel-arena-kernelbench-hip-source-min-run · How It Works · Data refreshed daily, snapshot 2026-10-09.