Beyond Pass Rate: leaderboard
Beyond Pass Rate (arXiv) — execution-grounded multilingual evaluation of open-source code LLMs that looks past raw pass@1 at answer correctness; mean correctness (%).
Metric: Correctness (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Yi-Coder-9B | 23.6 | #1428 |
| 2 | Qwen 2.5 14B | 22.2 | #725 |
| 3 | Gemma 2 27B | 18.9 | #806 |
| 4 | Phi-4 | 10.3 | #706 |
| 5 | Llama 3.1 8B | 8.1 | #1151 |
| 6 | granite-3.1-8B-instruct | 2.6 | #1069 |
| 7 | Stable-Code 3B | 0.7 | #1595 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=beyond-pass-rate · How It Works · Data refreshed daily, snapshot 2026-10-20.