Chess Puzzles
Rank #2 of 51 · 44 models scored · top score 64
unknown
Strongest on discrimination, weakest on saturation headroom. Caveat: the public test set leaves contamination on the table.
Weighted composite. Each signal is computed from the source-backed leaderboard below.
Does it still separate models?
How far from ceiling / clustered at the top?
Public vs held-out; training-leak risk.
Apples-to-apples, or mixed / vendor-optimized?
How old is the benchmark?
| # | Model | Score | Evidence | Measured |
|---|---|---|---|---|
| 1 | GPT-5.5 Pro | 64 | reproduced | 2026-04-23 |
| 2 | GPT-5.4 Pro | 58.6 | reproduced | 2026-03-05 |
| 3 | Gemini 3.1 Pro | 55 | reproduced | 2026-02-19 |
| 4 | GPT-5.5 | 54 | reproduced | 2026-04-23 |
| 5 | Gemini 3.5 Flash | 50 | reproduced | 2026-05-19 |
| 6 | GPT-5.2 | 49 | reproduced | 2025-12-11 |
| 7 | GPT-5.4 | 44 | reproduced | 2026-03-05 |
| 8 | Claude Fable 5 | 41 | reproduced | 2026-06-09 |
| 9 | Gemini 3 Flash | 38 | reproduced | 2025-12-17 |
| 10 | GPT-5 | 37 | reproduced | 2025-08-07 |
| 11 | Claude Opus 4.8 | 34 | reproduced | 2026-05-28 |
| 12 | GPT-5.1 | 32 | reproduced | 2025-11-13 |
| 13 | Gemini 3 Pro | 31 | reproduced | 2025-11-18 |
| 14 | Claude Opus 4.7 | 30 | reproduced | 2026-04-16 |
| 15 | GPT-5.4 Nano | 30 | reproduced | 2026-03-17 |
| 16 | Grok 4 | 28 | reproduced | 2025-07-09 |
| 17 | Kimi K2.6 | 26 | reproduced | 2026-04-20 |
| 18 | o4-mini | 26 | reproduced | 2025-04-16 |
| 19 | Grok 4.3 Beta | 25 | reproduced | 2026-04-17 |
| 20 | Qwen3.7-Max | 22 | reproduced | 2026-05-19 |
| 21 | Kimi K2.7 Code | 21 | reproduced | 2026-06-12 |
| 22 | GLM-5.2 | 21 | reproduced | 2026-06-16 |
| 23 | Kimi K2 Thinking | 20 | reproduced | 2025-11-06 |
| 24 | Qwen 3.5 Flash (hosted 35B-A3B) | 20 | reproduced | 2026-02-25 |
| 25 | gpt-oss-120b | 20 | reproduced | 2025-08-05 |
| 26 | DeepSeek-V4-Pro | 20 | reproduced | 2026-04-24 |
| 27 | Gemini 2.5 Pro (Jun 2025) | 20 | reproduced | 2025-06-05 |
| 28 | GLM-5.1 | 18 | reproduced | 2026-04-07 |
| 29 | GPT-5.4 Mini | 18 | reproduced | 2026-03-17 |
| 30 | Qwen 3.6 Flash | 17.17 | reproduced | 2026-04-26 |
| 31 | Qwen 3.5 Plus (hosted 397B-A17B) | 17 | reproduced | 2026-02-16 |
| 32 | Qwen 3.6 Max (Preview) | 17 | reproduced | 2026-04-20 |
| 33 | o3-mini | 17 | reproduced | 2025-01-31 |
| 34 | Claude Opus 4.6 | 17 | reproduced | 2026-02-05 |
| 35 | Qwen 3.6 Plus | 16.16 | reproduced | 2026-04-01 |
| 36 | DeepSeek-V3.2 | 14 | reproduced | 2025-12-01 |
| 37 | Claude Sonnet 4.6 | 13 | reproduced | 2026-02-17 |
| 38 | Kimi K2.5 | 12 | reproduced | 2026-02-02 |
| 39 | Qwen3-235B-A22B-Thinking (Jul 2025) | 12 | reproduced | 2025-07-25 |
| 40 | Claude Opus 4.5 | 12 | reproduced | 2025-11-24 |
| 41 | Claude Sonnet 4.5 | 12 | reproduced | 2025-09-29 |
| 42 | GLM-5 | 10 | reproduced | 2026-02-11 |
| 43 | GLM-4.7 | 6 | reproduced | 2025-12-22 |
| 44 | Qwen3-Max | 4 | reproduced | 2025-09-05 |