Benchmarks · coding
Aider Polyglot
Evaluates LLMs on 225 Exercism coding exercises across six languages, testing code editing without human intervention. Maintained by Paul Gauthier. Score: Percent of the 225 Exercism exercises completed with a correct edit.
37 models · 42 results · data as of 2026-10-29 · aider.chat ↗
Leaderboard
GPT-5 high
All results
42 rows| # | Model | Effort | Aider Polyglot | Harness |
|---|---|---|---|---|
| 1 | GPT-5 | high | 88.0% | |
| 2 | GPT-5 | medium | 86.7% | |
| 3 | o3 pro | high | 84.9% | |
| 4 | Gemini 2.5 Pro | high | 83.1% | |
| 5 | GPT-5 | low | 81.3% | |
| 6 | o3 | high | 81.3% | |
| 7 | Grok 4 | high | 79.6% | |
| 8 | Gemini 2.5 Pro | medium | 79.1% | |
| 9 | GPT-4.1 | high | 78.2% | |
| 10 | DeepSeek V3.2 | high | 74.2% | |
| 11 | Opus 4 | medium | 72.0% | |
| 12 | o4 mini | high | 72.0% | |
| 13 | DeepSeek R1 | xhigh | 71.4% | |
| 14 | o1 | high | 61.7% | |
| 15 | Sonnet 4 | high | 61.3% | |
| 16 | o3 mini | high | 60.4% | |
| 17 | Qwen 3 235B-A22B | xhigh | 59.6% | |
| 18 | Kimi K2 | medium | 59.1% | |
| 19 | DeepSeek V3 | none | 55.1% | aider, 'diff' edit format |
| 20 | Gemini 2.5 Flash | medium | 55.1% | |
| 21 | o3 mini | medium | 53.8% | |
| 22 | Grok 3 | none | 53.3% | |
| 23 | Sonnet 3.5 | medium | 51.6% | |
| 24 | Grok 3 Mini | high | 49.3% | |
| 25 | GPT-4.5 | 44.9% | ||
| 26 | gpt-oss-120b | high | 41.8% | |
| 27 | Qwen 3 32B | none | 40.0% | |
| 28 | Gemini 1206 | 38.2% | ||
| 29 | Gemini 2.0 Pro | none | 35.6% | |
| 30 | Grok 3 Mini | low | 34.7% |