Leaderboard

Frontier AI models ranked by chess understanding and puzzle solving ability.

Last updated: February 23, 2026
* Partial run: Models marked with * were tested on Semantic only (400 questions). Full 900-question runs include both Semantic and Position Judgment categories.
#ModelProviderUnderstandingPuzzle SolvingPuzzle ELOCost / 1K Q
1
Claude Opus 4.6400Q semantic only
Anthropic83.5%*Coming SoonComing Soon$20.10
2
GPT-5.2400Q semantic only
OpenAI78.8%*Coming SoonComing Soon$25.48
3
Claude Sonnet 4.5400Q semantic only
Anthropic73.0%*Coming SoonComing Soon$9.87
4
Gemini 3 Flash
Google58.6%Coming SoonComing Soon$1.87
5
GPT-5 Nano
OpenAI33.0%Coming SoonComing Soon$0.69

Cost vs. Understanding

How much chess knowledge do you get per dollar?