ChessBench - Accuracy: leaderboard
ChessBench move-quality score: how close a model’s chosen moves are to best play, averaged across branching, converting and resisting positions.
Metric: Accuracy (%). Source: chessbench.ai. Saturation forecast: Around December 2027. 72 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra | 85 | #6 |
| 2 | Gemini 4 Argon | 81.8 | #9 |
| 3 | Gemini 3.8 Flash | 81.1 | #30 |
| 4 | GPT-6.1 Sol | 80.7 | #7 |
| 5 | O3 | 80.5 | #132 |
| 6 | Muse Spark 1.1 | 79.9 | #55 |
| 7 | Gemini 3.1 Pro (Preview) | 79.4 | #64 |
| 8 | Claude Opus 5.5 | 79.3 | #3 |
| 9 | GPT-5.5 | 78.8 | #33 |
| 10 | Gemini 3.5 Flash | 78.7 | #66 |
| 11 | GPT-6.1 Terra | 78.3 | #15 |
| 12 | Gemini 3.7 Flash | 78.1 | #24 |
| 13 | GPT-6 Sol | 77.4 | #19 |
| 14 | Claude Fable 5 | 77.1 | #20 |
| 15 | GPT-5.4 | 76.5 | #83 |
Interactive version: theaggregate.ai/benchmark?slug=chessbench-accuracy · How It Works · Data refreshed daily, snapshot 2026-11-05.