Aider Code Editing Leaderboard: leaderboard
Aider's legacy code editing benchmark: 133 Exercism Python exercises solved through Aider edit formats, scored as the percentage completed correctly after one retry with unit-test feedback (pass_rate_2); higher is better. Frozen since December 2024; incomplete and mis-filed runs are excluded.
Metric: Exercises completed correctly after one retry, pass_rate_2 (%). Source: aider.chat. Saturation forecast: Estimated already saturated. 94 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.5 Sonnet (20241022) | 84.2 | #302 |
| 2 | O1 | 84.2 | #190 |
| 3 | O1 Preview | 79.7 | #205 |
| 4 | Claude 3.5 Sonnet (20240620) | 77.4 | #308 |
| 5 | Claude 3.5 Haiku (20241022) | 75.2 | #590 |
| 6 | GPT-4o (2024-05-13) | 72.9 | #337 |
| 7 | GPT-4o ChatGPT | 72.2 | |
| 8 | DeepSeek V2.5 | 72.2 | #494 |
| 9 | GPT-4o (2024-08-06) | 71.4 | #343 |
| 10 | GPT-4o (2024-11-20) | 71.4 | #392 |
| 11 | Qwen 2.5 Coder 32B Instruct | 71.4 | #534 |
| 12 | O1 Mini | 70.7 | #334 |
| 13 | Gemini 2.0 Flash (Preview) | 69.9 | #376 |
| 14 | Qwen 2.5 Coder 14B Instruct | 69.2 | #685 |
| 15 | Claude 3 Opus (20240229) | 68.4 | #428 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=aider-code-editing-leaderboard · How It Works · Data refreshed daily, snapshot 2026-11-05.