ParallelKernelBench Pass@1: leaderboard
ParallelKernelBench pass@1 measures single-shot correctness on production-style multi-GPU kernel generation tasks.
Metric: Pass@1 (% of problems). Source: www.together.ai. Saturation forecast: Around June 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 32.18 | #31 |
| 2 | Gemini 3 Pro | 27.59 | #80 |
| 3 | Claude Opus 4.7 | 22.99 | #49 |
| 4 | GLM-5.1 | 6.9 | #97 |
| 5 | GLM-5.2 | 4.6 | #73 |
| 6 | DeepSeek V4 Pro | 2.3 | #102 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=parallelkernelbench-pass-1 · How It Works · Data refreshed daily, snapshot 2026-10-20.