Software Issue Resolution on SWE-rebench 60-task Python subset v2
36.11Pass@1Claude Opus-4.5
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Claude Opus-4.5Evaluation Harness=mini-swe-agent harness2026.05 | 36.11 | 36.67 | |
| GLM-4.7Evaluation Harness=mini-swe-agent harness2026.05 | 27.22 | 31.67 | |
| MiniMax-M2.1Evaluation Harness=mini-swe-agent harness2026.05 | 26.11 | 31.67 | |
| GeminiEvaluation Harness=mini-swe-agent harness2026.05 | 25.56 | 33.33 | |
| DeepSeek-V3.2Evaluation Harness=mini-swe-agent harness2026.05 | 23.33 | 31.67 | |
| GPT-5.2Evaluation Harness=mini-swe-agent harness2026.05 | 20.56 | 23.33 | |
| gpt-oss-120bEvaluation Harness=mini-swe-agent harness2026.05 | 8.89 | 16.67 |