Sphinx - Python: leaderboard
Sphinx checklist-coverage score on its 500 Python pull requests, 450 needing changes and 50 bug-free controls whose correct review is no comment: a GPT-4o judge checks each generated review against a checklist derived from the merged change, buggy cases weighted 0.9; percent, higher is better. Microsoft and University of Rochester, 2026.
Metric: Checklist Coverage Score (%; GPT-4o judge, 450 buggy + 50 bug-free PRs per language). Source: arxiv.org. Saturation forecast: Estimated already saturated. 29 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4.1 | 42.4 | #492 |
| 2 | O4 Mini | 40.81 | #354 |
| 3 | Gemini 2.5 Pro | 40.8 | #262 |
| 4 | O3 Mini | 39.73 | #536 |
| 5 | Qwen 3 235B A22B | 38.88 | #622 |
| 6 | Claude 3.7 Sonnet | 38.2 | #455 |
| 7 | Qwen 2.5 72B Instruct | 38.13 | #1093 |
| 8 | Qwen 3 30B A3B | 38.08 | #1124 |
| 9 | GPT-4o (2024-08-06) | 37.15 | #731 |
| 10 | Claude 3.5 Sonnet | 36.95 | #747 |
| 11 | GPT-4o Mini | 36.73 | #1627 |
| 12 | DeepSeek V3 | 32.99 | #705 |
| 13 | GPT-4.1 Mini | 32.78 | #775 |
| 14 | DeepSeek R1 | 32.6 | #467 |
| 15 | Qwen 3 32B | 32.15 | #960 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=sphinx-python · How It Works · Data refreshed daily, snapshot 2026-11-12.