A harder extension of SWE-bench focusing on more complex, professional-grade software engineering tasks from real GitHub issues requiring multi-file changes and deeper codebase understanding.
This definition has not completed editorial and provenance review. Verify its metric, version, source, and evaluation configuration upstream before using it for a model-selection decision.
Tests professional software engineering beyond simple function generation. More challenging than SWE-bench Verified.
Newer benchmark with fewer baselines. Score range (26–57%) shows room for improvement.
Higher is better. Percentage of issues resolved (0–100%).
Newer benchmark with fewer baselines.
| # | Model | Score |
|---|---|---|
| 1 | 90.3 | |
| 2 | 87.4 | |
| 3 | 86.9 | |
| 4 | 81.7 | |
| 5 | 81.7 | |
| 6 | 80.7 | |
| 7 | 78.3 | |
| 8 | 77.8 | |
| 9 | 67.1 | |
| 10 | 65.8 | |
| 11 | 65.3 | |
| 12 | 64.2 | |
| 13 | 63.6 | |
| 14 | 62.7 | |
| 15 | 62.3 | |
| 16 | 61.8 | |
| 17 | 60.5 | |
| 18 | 60.1 | |
| 19 | 59.5 | |
| 20 | 59.1 | |
| 21 | 59.1 | |
| 22 | 58.9 | |
| 23 | 58.8 | |
| 24 | 58.7 | |
| 25 | 58.7 | |
| 26 | 58.6 | |
| 27 | 58.6 | |
| 28 | 58.0 | |
| 29 | 57.9 | |
| 30 | 57.7 | |
| 31 | 57.6 | |
| 32 | 57.0 | |
| 33 | 56.8 | |
| 34 | 56.8 | |
| 35 | 56.3 | |
| 36 | 56.0 | |
| 37 | 56.0 | |
| 38 | 55.7 | |
| 39 | 55.7 | |
| 40 | 55.7 | |
| 41 | 55.7 | |
| 42 | 55.6 | |
| 43 | 55.6 | |
| 44 | 55.6 | |
| 45 | 55.6 | |
| 46 | 55.5 | |
| 47 | 55.3 | |
| 48 | 55.1 | |
| 49 | 54.6 | |
| 50 | 54.4 |