SWE-Bench ProMax - Python (OpenHands): leaderboard
The 29 Python instances of SWE-Bench ProMax (2026), expert-curated refactoring tasks from real commits that are resolved when the patch passes every test of the reviewed suite. Resolve rate in percent with the OpenHands scaffold; higher is better.
Metric: Resolve rate (%, Pass@1 on the 29 Python instances; 170 repository-level refactoring instances from real commits (average 11.4 modified files), resolved when the patch passes every test of the manually reviewed suite; OpenHands scaffold, step limit 300, cost limit 10 US dollars per instance). Source: arxiv.org. Saturation forecast: Not forecast. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 | 48.3 | #176 |
| 2 | Kimi K2.5 | 24.1 | #256 |
| 3 | GLM-5 | 20.7 | #260 |
| 4 | Claude Sonnet 4.6 | 17.2 | #161 |
| 5 | Gemini 3 Pro | 13.8 | #137 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=swe-bench-promax-python-openhands · How It Works · Data refreshed daily, snapshot 2026-11-13.