ChessBench: leaderboard
Chess benchmark where language models play rated games against fixed reference opponents, scored on rule-compliant play, move quality, and Monte Carlo Elo.
Metric: Overall score (Stockfish 19 = 100). Source: chessbench.ai. Saturation forecast: Around January 2028. 72 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra | 36.36 | #6 |
| 2 | Gemini 3.8 Flash | 29.32 | #30 |
| 3 | Gemini 4 Argon | 28.54 | #9 |
| 4 | GPT-6.1 Sol | 27.92 | #7 |
| 5 | GPT-5.5 | 27.06 | #33 |
| 6 | Claude Opus 5.5 | 22.65 | #3 |
| 7 | GPT-6.1 Terra | 22.59 | #15 |
| 8 | GPT-5.6 Sol | 22.24 | #22 |
| 9 | Gemini 3.7 Flash | 21.8 | #24 |
| 10 | GPT-6 Sol | 21.73 | #19 |
| 11 | Gemini 3.1 Pro (Preview) | 21.27 | #64 |
| 12 | Gemini 3.6 Flash | 20.76 | #57 |
| 13 | Gemini 3.5 Flash | 19.03 | #66 |
| 14 | Muse Spark 1.3 | 18.77 | #31 |
| 15 | Claude Fable 5 | 18.36 | #20 |
Interactive version: theaggregate.ai/benchmark?slug=chessbench · How It Works · Data refreshed daily, snapshot 2026-11-05.