Benchmarks
Chess Puzzles
Models scored
85
Best score
0.64
Metric
mean_score
Chess Puzzles is an AI evaluation tracked by Epoch AI, with 85 scored model versions on record. Epoch AI runs this evaluation directly, so the scores are reproducible against their published logs.
Scored models span February 2024 to July 2026. The best score recorded is 0.64, measured as mean_score. Models come from Anthropic, Moonshot, DeepSeek, OpenAI, Google, xAI, Z.ai (Zhipu AI), Alibaba and others.
Full record
- Score metric
- mean_score
- Models scored
- 85
- Best score recorded
- 0.64
- Earliest model scored
- Feb 29, 2024
- Most recent model scored
- Jul 24, 2026
- Third-party reported
- No
Leaderboard
| 01 | gpt-5.5-pro-pre-release_xhigh | OpenAI | 0.64 |
| 02 | gpt-5.6-sol_promax | OpenAI | 0.64 |
| 03 | gpt-5.4-pro-2026-03-05_xhigh | OpenAI | 0.59 |
| 04 | gemini-3.1-pro-preview | Google DeepMind | 0.55 |
| 05 | gpt-5.6-sol_max | OpenAI | 0.55 |
| 06 | gpt-5.5-pre-release_xhigh | OpenAI | 0.54 |
| 07 | gpt-5.6-terra_max | OpenAI | 0.54 |
| 08 | gemini-3.5-flash_high | 0.5 | |
| 09 | gpt-5.2-2025-12-11_xhigh | OpenAI | 0.49 |
| 10 | gpt-5.4-2026-03-05_xhigh | OpenAI | 0.44 |
| 11 | gemini-3.5-flash_minimal | 0.43 | |
| 12 | claude-opus-5_max | Anthropic | 0.42 |
| 13 | claude-fable-5_max | Anthropic | 0.41 |
| 14 | gpt-5.2-2025-12-11_high | OpenAI | 0.4 |
| 15 | gpt-5.6-luna_max | OpenAI | 0.4 |
| 16 | gpt-5.2-2025-12-11_medium | OpenAI | 0.4 |
| 17 | kimi-k3_max | Moonshot | 0.39 |
| 18 | gemini-3-flash-preview | Google DeepMind | 0.38 |
| 19 | gpt-5.4-2026-03-05_medium | OpenAI | 0.38 |
| 20 | gpt-5.4-2026-03-05_high | OpenAI | 0.38 |
| 21 | gpt-5-2025-08-07_high | OpenAI | 0.37 |
| 22 | grok-4.5_high | xAI | 0.36 |
| 23 | claude-sonnet-5_xhigh | Anthropic | 0.35 |
| 24 | claude-opus-4-8_max | Anthropic | 0.34 |
| 25 | gpt-5.1-2025-11-13_high | OpenAI | 0.32 |
| 26 | gemini-3-pro-preview | Google DeepMind | 0.31 |
| 27 | gpt-5.4-nano-2026-03-17_high | OpenAI | 0.3 |
| 28 | claude-opus-4-7_xhigh | Anthropic | 0.3 |
| 29 | gpt-5-2025-08-07_medium | OpenAI | 0.29 |
| 30 | grok-4-0709 | xAI | 0.28 |
| 31 | o3-2025-04-16_low | OpenAI | 0.27 |
| 32 | kimi-k2.6 | Moonshot | 0.26 |
| 33 | o4-mini-2025-04-16_high | OpenAI | 0.26 |
| 34 | grok-4.3_high | xAI | 0.25 |
| 35 | gpt-5-2025-08-07_low | OpenAI | 0.24 |
| 36 | grok-4.20-0309-reasoning | xAI | 0.24 |
| 37 | gpt-5.2-2025-12-11_low | OpenAI | 0.23 |
| 38 | qwen3.7-max | Alibaba | 0.22 |
| 39 | glm-5.2_max | Z.ai (Zhipu AI) | 0.21 |
| 40 | kimi-k2.7-code | Moonshot | 0.21 |
| 41 | openai/gpt-oss-120b_high | OpenAI | 0.2 |
| 42 | qwen3.5-flash | Alibaba | 0.2 |
| 43 | kimi-k2-thinking-turbo | Moonshot | 0.2 |
| 44 | gpt-5.4-2026-03-05_low | OpenAI | 0.2 |
| 45 | gemini-2.5-pro | Google DeepMind | 0.2 |
| 46 | claude-opus-4-7_low | Anthropic | 0.2 |
| 47 | deepseek-v4-pro_max | DeepSeek | 0.2 |
| 48 | gpt-5.4-mini-2026-03-17_high | OpenAI | 0.18 |
| 49 | glm-5.1 | Z.ai (Zhipu AI) | 0.18 |
| 50 | qwen3.6-flash | Alibaba | 0.17 |
| 51 | claude-opus-4-6_32K | Anthropic | 0.17 |
| 52 | o3-mini-2025-01-31_high | OpenAI | 0.17 |
| 53 | qwen3.6-max-preview | Alibaba | 0.17 |
| 54 | qwen3.5-plus | Alibaba | 0.17 |
| 55 | qwen3.6-plus | Alibaba | 0.16 |
| 56 | claude-sonnet-5_max | Anthropic | 0.16 |
| 57 | gpt-5-2025-08-07_minimal | OpenAI | 0.16 |
| 58 | gpt-5-nano-2025-08-07_low | OpenAI | 0.15 |
| 59 | o4-mini-2025-04-16_low | OpenAI | 0.14 |
| 60 | deepseek-reasoner | DeepSeek | 0.14 |
More benchmarks
SourceEpoch AI, 'AI Benchmarking Hub'. Published online at epoch.ai. Retrieved 2026-07-29 from https://epoch.ai/benchmarks. Licensed under CC BY 4.0.
← All benchmarks