# Chess Puzzles (Epoch AI) (Chess Puzzles)

> Epoch AI's Chess Puzzles benchmark gives models 100 novel chess positions as FEN text and asks for the single best next move, as judged by Stockfish.

Canonical page: https://benchlm.ai/benchmarks/chesspuzzles

- Category: [Reasoning](/reasoning)
- Last updated: November 13, 2026

## About Chess Puzzles

- Year: 2026
- Tasks: 100 programmatically generated chess positions with a single best move
- Format: Exact-match move answer from FEN text input
- Difficulty: Novel chess positions, single best move
- Paper: [Chess Puzzles (Epoch AI)](https://epoch.ai/benchmarks/chess-puzzles)

Epoch generated every position programmatically, so none appear in any other source. Games open with six random moves, continue with Stockfish at a random Elo rating, and full-strength Stockfish then picks positions with exactly one best move. Positions where the best move is an obvious capture are filtered out. Models get the board as a FEN string (no image), a budget of up to 1,000,000 output tokens, and must submit a move such as b1c3 through a tool call. The answer must match the key exactly.

Chess Puzzles is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (81 models)

| Rank | Model | Creator | Score | Who ran it |
|------|-------|---------|-------|-----------|
| 1 | [Claude Fable 5.2](/models/claude-fable-5-2) | Anthropic | 79.0% | Run by Chess Puzzles |
| 2 | [Gemini 4 Carbon](/models/gemini-4-carbon) | Google | 78.0% | Run by Chess Puzzles |
| 3 | [GPT-6 Astra](/models/gpt-6-astra) | OpenAI | 74.0% | Run by Chess Puzzles |
| 4 | [GPT-6.1 Sol](/models/gpt-6-1-sol) | OpenAI | 68.0% | Run by Chess Puzzles |
| 5 | [Claude Opus 5.5](/models/claude-opus-5-5) | Anthropic | 67.0% | Run by Chess Puzzles |
| 6 | [Gemini 3.7 Flash](/models/gemini-3-7-flash) | Google | 66.0% | Run by Chess Puzzles |
| 7 | [GPT-5.5](/models/gpt-5-5) | OpenAI | 63.0% | Run by Chess Puzzles |
| 8 | [GPT-5.5 Pro](/models/gpt-5-5-pro) | OpenAI | 62.0% | Run by Chess Puzzles |
| 9 | [Claude Sonnet 5.5](/models/claude-sonnet-5-5) | Anthropic | 59.0% | Run by Chess Puzzles |
| 10 | [Gemini 4 Argon](/models/gemini-4-argon) | Google | 59.0% | Run by Chess Puzzles |
| 11 | [GPT-5.6 Sol](/models/gpt-5-6-sol) | OpenAI | 59.0% | Run by Chess Puzzles |
| 12 | [GPT-5.4 Pro](/models/gpt-5-4-pro) | OpenAI | 59.0% | Run by Chess Puzzles |
| 13 | [Gemini 3.8 Flash](/models/gemini-3-8-flash) | Google | 58.0% | Run by Chess Puzzles |
| 14 | [GPT-6 Sol](/models/gpt-6-sol) | OpenAI | 56.0% | Run by Chess Puzzles |
| 15 | [Claude Haiku 5.5](/models/claude-haiku-5-5) | Anthropic | 55.0% | Run by Chess Puzzles |
| 16 | [Claude Opus 4.8](/models/claude-opus-4-8) | Anthropic | 52.0% | Run by Chess Puzzles |
| 17 | [Gemini 3.1 Pro](/models/gemini-3-1-pro) | Google | 51.0% | Run by Chess Puzzles |
| 18 | [Claude Fable 5.1](/models/claude-fable-5-1) | Anthropic | 48.0% | Run by Chess Puzzles |
| 19 | [GPT-5.4](/models/gpt-5-4) | OpenAI | 48.0% | Run by Chess Puzzles |
| 20 | [GPT-5.2](/models/gpt-5-2) | OpenAI | 48.0% | Run by Chess Puzzles |
| 21 | [Claude Fable 5](/models/claude-fable) | Anthropic | 47.0% | Run by Chess Puzzles |
| 22 | [GPT-5.6 Terra](/models/gpt-5-6-terra) | OpenAI | 47.0% | Run by Chess Puzzles |
| 23 | [DeepSeek V4.1 Flash 1017](/models/deepseek-v4-1-flash) | DeepSeek | 46.0% | Run by Chess Puzzles |
| 24 | [Qwen3.8 Max](/models/qwen3-8-max) | Alibaba | 46.0% | Run by Chess Puzzles |
| 25 | [DeepSeek V4 Pro 0813](/models/deepseek-v4-pro-0813) | DeepSeek | 46.0% | Run by Chess Puzzles |
| 26 | [Kimi K3.1 Pro](/models/kimi-3-1-pro) | Moonshot AI | 45.0% | Run by Chess Puzzles |
| 27 | [Claude Sonnet 5](/models/claude-sonnet-5) | Anthropic | 45.0% | Run by Chess Puzzles |
| 28 | [Gemini 3.5 Flash](/models/gemini-3-5-flash) | Google | 45.0% | Run by Chess Puzzles |
| 29 | [GPT-5.4 nano](/models/gpt-5-4-nano) | OpenAI | 45.0% | Run by Chess Puzzles |
| 30 | [Claude Opus 5](/models/claude-opus-5) | Anthropic | 43.0% | Run by Chess Puzzles |
| 31 | [GPT-5.6 Luna](/models/gpt-5-6-luna) | OpenAI | 43.0% | Run by Chess Puzzles |
| 32 | [GPT-5.1](/models/gpt-5-1) | OpenAI | 42.0% | Run by Chess Puzzles |
| 33 | [Claude Opus 4.7](/models/claude-opus-4-7) | Anthropic | 41.0% | Run by Chess Puzzles |
| 34 | [Grok 4.7](/models/grok-4-7) | xAI | 41.0% | Run by Chess Puzzles |
| 35 | [Gemini 3.6 Flash](/models/gemini-3-6-flash) | Google | 40.0% | Run by Chess Puzzles |
| 36 | [Gemini 3 Flash](/models/gemini-3-flash) | Google | 40.0% | Run by Chess Puzzles |
| 37 | [GLM-5.3](/models/glm-5-3) | Z.AI | 37.0% | Run by Chess Puzzles |
| 38 | [Muse Spark 1.3](/models/muse-spark-1-3) | Meta | 37.0% | Run by Chess Puzzles |
| 39 | [Kimi K3](/models/kimi-k3) | Moonshot AI | 36.0% | Run by Chess Puzzles |
| 40 | [GPT-6 Luna](/models/gpt-6-luna) | OpenAI | 36.0% | Run by Chess Puzzles |
| 41 | [o3](/models/o3) | OpenAI | 36.0% | Run by Chess Puzzles |
| 42 | [GPT-5 (high)](/models/gpt-5-high) | OpenAI | 36.0% | Run by Chess Puzzles |
| 43 | [Grok 4.5](/models/grok-4-5) | xAI | 35.0% | Run by Chess Puzzles |
| 44 | [Muse Spark 1.1](/models/muse-spark-1-1) | Meta | 34.0% | Run by Chess Puzzles |
| 45 | [Grok 4.6](/models/grok-4-6) | xAI | 33.0% | Run by Chess Puzzles |
| 46 | [Mistral Large 4](/models/mistral-large-4) | Mistral | 33.0% | Run by Chess Puzzles |
| 47 | [GPT-5 (medium)](/models/gpt-5-medium) | OpenAI | 32.0% | Run by Chess Puzzles |
| 48 | [DeepSeek V4 Flash 0731](/models/deepseek-v4-flash-0731) | DeepSeek | 32.0% | Run by Chess Puzzles |
| 49 | [Grok 4.20](/models/grok-4-20-beta) | xAI | 29.0% | Run by Chess Puzzles |
| 50 | [Muse Spark 1.2](/models/muse-spark-1-2) | Meta | 27.0% | Run by Chess Puzzles |
| 51 | [Qwen3.7 Max](/models/qwen3-7-max) | Alibaba | 27.0% | Run by Chess Puzzles |
| 52 | [Seed 2.1 Turbo](/models/seed-2-1-turbo) | ByteDance | 26.0% | Run by Chess Puzzles |
| 53 | [Claude Sonnet 4.6](/models/claude-sonnet-4-6) | Anthropic | 23.0% | Run by Chess Puzzles |
| 54 | [GLM-5.2](/models/glm-5-2) | Z.AI | 22.0% | Run by Chess Puzzles |
| 55 | [Claude Opus 4.6](/models/claude-opus-4-6) | Anthropic | 21.0% | Run by Chess Puzzles |
| 56 | [GLM-5.3-Flash](/models/glm-5-3-flash) | Z.AI | 21.0% | Run by Chess Puzzles |
| 57 | [Inkling](/models/inkling) | Thinking Machines Lab | 21.0% | Run by Chess Puzzles |
| 58 | [GLM-4.7](/models/glm-4-7) | Z.AI | 21.0% | Run by Chess Puzzles |
| 59 | [Seed 2.1 Pro](/models/seed-2-1-pro) | ByteDance | 21.0% | Run by Chess Puzzles |
| 60 | [Kimi K2.6](/models/kimi-2-6) | Moonshot AI | 20.0% | Run by Chess Puzzles |
| 61 | [GLM-5.1](/models/glm-5-1) | Z.AI | 20.0% | Run by Chess Puzzles |
| 62 | [Grok 4.3](/models/grok-4-3) | xAI | 18.0% | Run by Chess Puzzles |
| 63 | [GPT-5.4 mini](/models/gpt-5-4-mini) | OpenAI | 17.0% | Run by Chess Puzzles |
| 64 | [MiniMax M3](/models/minimax-m3) | MiniMax | 17.0% | Run by Chess Puzzles |
| 65 | [Qwen3.6 Plus](/models/qwen3-6-plus) | Alibaba | 16.0% | Run by Chess Puzzles |
| 66 | [Gemini 3.5 Flash-Lite](/models/gemini-3-5-flash-lite) | Google | 16.0% | Run by Chess Puzzles |
| 67 | [o3-mini](/models/o3-mini) | OpenAI | 15.0% | Run by Chess Puzzles |
| 68 | [Claude Opus 4.5](/models/claude-opus-4-5) | Anthropic | 14.0% | Run by Chess Puzzles |
| 69 | [Kimi K2.7 Code](/models/kimi-k2-7-code) | Moonshot AI | 14.0% | Run by Chess Puzzles |
| 70 | [Qwen3.5 Flash](/models/qwen3-5-flash) | Alibaba | 14.0% | Run by Chess Puzzles |
| 71 | [o4-mini (high)](/models/o4-mini-high) | OpenAI | 13.0% | Run by Chess Puzzles |
| 72 | [GPT-4.1](/models/gpt-4-1) | OpenAI | 11.0% | Run by Chess Puzzles |
| 73 | [Claude Sonnet 4.5](/models/claude-sonnet-4-5) | Anthropic | 10.0% | Run by Chess Puzzles |
| 74 | [o1](/models/o1) | OpenAI | 10.0% | Run by Chess Puzzles |
| 75 | [Qwen3.5 Plus](/models/qwen3-5-plus) | Alibaba | 9.0% | Run by Chess Puzzles |
| 76 | [GPT-4o](/models/gpt-4o) | OpenAI | 9.0% | Run by Chess Puzzles |
| 77 | [Claude Haiku 4.5](/models/claude-haiku-4-5) | Anthropic | 8.0% | Run by Chess Puzzles |
| 78 | [Claude 3 Opus](/models/claude-3-opus) | Anthropic | 4.0% | Run by Chess Puzzles |
| 79 | [GPT-4o mini](/models/gpt-4o-mini) | OpenAI | 3.0% | Run by Chess Puzzles |
| 80 | [GPT-4.1 mini](/models/gpt-4-1-mini) | OpenAI | 1.0% | Run by Chess Puzzles |
| 81 | [GPT-4.1 nano](/models/gpt-4-1-nano) | OpenAI | 0.0% | Run by Chess Puzzles |

## FAQ

### What does the Chess Puzzles benchmark test?

Whether a model can find the one best next move in a chess position it has never seen. Epoch generated the positions itself, so they cannot be memorized from puzzle sites.

### Which model is best at chess puzzles?

Claude Fable 5.2 at max effort leads with 79%, followed by GPT-6 Astra at max (74%), GPT-6.1 Sol (68%) and Claude Opus 5.5 (67%).

## Compare Top Models on Chess Puzzles

- [Claude Fable 5.2 vs Gemini 4 Carbon](/compare/claude-fable-5-2-vs-gemini-4-carbon)
- [Gemini 4 Carbon vs GPT-6 Astra](/compare/gemini-4-carbon-vs-gpt-6-astra)
- [GPT-6 Astra vs GPT-6.1 Sol](/compare/gpt-6-1-sol-vs-gpt-6-astra)
- [GPT-6.1 Sol vs Claude Opus 5.5](/compare/claude-opus-5-5-vs-gpt-6-1-sol)