SWE-bench Verified (SWE-ABS Tests): leaderboard
Metric: Resolve rate (%) on the 500 SWE-bench Verified tasks when each bash-only (mini-SWE-agent) leaderboard submission's patches are re-run against the SWE-ABS strengthened test suites (original tests plus coverage- and mutation-guided tests, overfitting tests fixed); higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 29 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 (2025-12-11) (High) | 56.8 | |
| 2 | Claude Sonnet 4.5 | 54.4 | #138 |
| 3 | GPT-5.2 | 51.6 | #105 |
| 4 | Claude Opus 4 (20250514) | 50.6 | #148 |
| 5 | GPT-5 (Medium) | 48.6 | #91 (GPT-5) |
| 6 | Claude Sonnet 4 (20250514) | 45.8 | #211 |
| 7 | Kimi K2 (Thinking) | 45 | #236 (Kimi K2) |
| 8 | GPT-5 Mini (2025-08-07) (Medium) | 44.6 | #165 (GPT-5 Mini (2025-08-07)) |
| 9 | MiniMax-M2 | 44.4 | #307 |
| 10 | O3 (2025-04-16) | 43.8 | #117 |
| 11 | Gemini 2.5 Pro (Preview 05-06) | 40.8 | #143 |
| 12 | GLM-4.6 | 40.2 | #246 |
| 13 | GLM-4.5 | 38.2 | #265 |
| 14 | GPT-5 Nano (2025-08-07) (Medium) | 22.8 | |
| 15 | GPT-OSS-120B | 20.2 | #330 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=swe-bench-verified-swe-abs-tests · How It Works · Data refreshed daily, snapshot 2026-10-11.