Benchmarks · reasoning
Chess Puzzles (Epoch AI)
Epoch AI's chess puzzles: the model finds the best move in positions with a single correct line. Score is the share of puzzles solved. Data: Epoch AI, CC BY.
84 models · 148 results · data as of 2026-10-29 · epoch.ai ↗
Leaderboard
GPT-6 Astra Pro max
All results
148 rows| # | Model | Effort | Chess Puzzles (Epoch AI) | Harness |
|---|---|---|---|---|
| 1 | GPT-6 Astra Pro | max | 76.0% | Epoch AI |
| 2 | GPT-6 Astra | max | 74.0% | Epoch AI |
| 3 | GPT-5.6 Sol Pro | max | 68.0% | Epoch AI |
| 4 | GPT-6.1 Sol | max | 68.0% | Epoch AI |
| 5 | Opus 5.5 | max | 67.0% | Epoch AI |
| 6 | Gemini 3.7 Flash | high | 66.0% | Epoch AI |
| 7 | GPT-5.5 | xhigh | 63.0% | Epoch AI |
| 8 | GPT-5.5 Pro | xhigh | 62.0% | Epoch AI |
| 9 | GPT-5.2 Pro | xhigh | 60.0% | Epoch AI |
| 10 | Sonnet 5.5 | max | 59.0% | Epoch AI |
| 11 | Gemini 4 Argon | high | 59.0% | Epoch AI |
| 12 | GPT-5.4 Pro | xhigh | 59.0% | Epoch AI |
| 13 | GPT-5.6 Sol | max | 59.0% | Epoch AI |
| 14 | Gemini 3.8 Flash | high | 58.0% | Epoch AI |
| 15 | GPT-6 Sol | max | 56.0% | Epoch AI |
| 16 | Haiku 5.5 | max | 55.0% | Epoch AI |
| 17 | Opus 4.8 | max | 52.0% | Epoch AI |
| 18 | Gemini 3.1 Pro | high | 51.0% | Epoch AI |
| 19 | Fable 5.1 | max | 48.0% | Epoch AI |
| 20 | DeepSeek V4.1 Flash | max | 48.0% | Epoch AI |
| 21 | GPT-5 Pro | high | 48.0% | Epoch AI |
| 22 | GPT-5.2 | xhigh | 48.0% | Epoch AI |
| 23 | GPT-5.4 | xhigh | 48.0% | Epoch AI |
| 24 | Fable 5 | max | 47.0% | Epoch AI |
| 25 | GPT-5.6 Terra | max | 47.0% | Epoch AI |
| 26 | DeepSeek V4 Pro 0813 | max | 46.0% | Epoch AI |
| 27 | DeepSeek V4.1 Flash 1017 | max | 46.0% | Epoch AI |
| 28 | Qwen 3.8 Max | xhigh | 46.0% | Epoch AI |
| 29 | Fable 5 | high | 45.0% | Epoch AI |
| 30 | Sonnet 5 | max | 45.0% | Epoch AI |