- Cleus
- AI benchmarks
- Chess Puzzles
Chess Puzzlesexplained
Find the one best move from a written-out position. The model gets 100 new chess positions written out as text (FEN notation), with no picture and no chess engine, and has to name the single best move. The positions were generated with the Stockfish engine from real-looking games where a player missed the one strong move.
- Highest published score
- GPT-6 Astra, 76.0%
- Leader among popular models
- GPT-6 Astra
- Models with a result
- 91
- Skill it tests
- Reasoning
What Chess Puzzles really tests
- What it is
- The model gets 100 new chess positions written out as text (FEN notation), with no picture and no chess engine, and has to name the single best move. The positions were generated with the Stockfish engine from real-looking games where a player missed the one strong move.
- A question like
- Given the position as a line of text, answer with one move like “e2e4”. Only the move Stockfish rates best counts. Our own example in the style of the test, not a real item.
- Why it matters
- It tests careful step-by-step look-ahead and keeping track of a precise state in its head, a skill that carries over to planning and debugging.
- Reading the score
- Share of the 100 puzzles where it named the best move. Five human players without aids, given 20 puzzles each, scored between 40% and 75%.
- The catch
- Chess is a narrow skill. A model can be great at it and average at everything else, or the reverse.
- Made by
- Epoch AI, 2025. Official page
Leaderboard
Each model’s best published result across its reasoning settings.
Popular models
The models we track across this site.
Top 20 of all 91 models tested
Includes older, preview and specialist models that aren’t on our main leaderboard.
Questions
What is Chess Puzzles?
The model gets 100 new chess positions written out as text (FEN notation), with no picture and no chess engine, and has to name the single best move. The positions were generated with the Stockfish engine from real-looking games where a player missed the one strong move.
What does a Chess Puzzles score mean?
Share of the 100 puzzles where it named the best move. Five human players without aids, given 20 puzzles each, scored between 40% and 75%.
Which AI model is best on Chess Puzzles?
GPT-6 Astra from OpenAI has the highest published Chess Puzzles result we have (76.0%).
Is Chess Puzzles a good benchmark?
It tests careful step-by-step look-ahead and keeping track of a precise state in its head, a skill that carries over to planning and debugging. The catch: Chess is a narrow skill. A model can be great at it and average at everything else, or the reverse.
Who made Chess Puzzles?
Epoch AI, 2025.
Benchmark scores: Epoch AI, Capabilities & benchmarking, used under CC BY 4.0. For each model we show its best published result across reasoning settings. Prices and context sizes: OpenRouter. Data refreshed November 9, 2026. Hands-on tests (answers, websites, code, refusals, speed) are our own single runs. Explanations written by Cleus.