ChessBench — Mate in 2, Historical Output Repair
Games · 2026-03-25
The 30-puzzle Mate in 2 track under ChessBench's historical protocol permitting one follow-up request after a missing or unparseable output; repaired exact UCI lines count as correct. Retained separately from single-request runs.
Top models (higher is better)
| Model | Score |
|---|---|
| Grok Build 0.1 | 86.7 |
| Gemini 3.6 Flash | 83.3 |
| Gemini 3.7 Flash | 83.3 |
| Grok 4.20 | 83.3 |
| GPT-6 Astra | 80.0 |
| Gemini 3.5 Flash | 76.7 |
| Grok 4.1 Fast | 73.3 |
| Grok 4.5 | 73.3 |
| GPT-5.6 Sol | 66.7 |
| Grok 4.3 | 63.3 |
| Gemini 3.1 Pro Preview | 56.7 |
| GPT-5.6 Terra | 53.3 |
| Claude Fable 5 | 43.3 |
| GPT-5.4 | 43.3 |
| Gemini 3.1 Flash Image Preview | 40.0 |
| Claude Opus 5 | 23.3 |
| GPT-5.5 | 20.0 |
| Opus 4.8 | 16.7 |
| Sonnet 5 | 16.7 |
| GPT-5.6 Luna | 16.7 |
| Qwen3.6 Plus (2026-04-02) | 16.7 |
| Qwen3.6 Plus Preview | 16.7 |
| Opus 4.6 | 10.0 |
| Opus 4.7 | 6.7 |
| Sonnet 4.6 | 3.3 |
| GLM-5.1 | 3.3 |
| GLM-5.2 | 3.3 |
| Gemini 2.5 Pro | 0.0 |
| GLM-5 | 0.0 |