ChessBench — Pin
Games · 2026-03-22
ChessBench Pin contains 30 fixed non-mating Lichess tactics labeled with the pin theme. Models are scored only when their returned UCI move line exactly matches the expected line; the track is evenly stratified across three puzzle-rating buckets.
Top models (higher is better)
| Model | Score |
|---|---|
| Grok 4.20 | 46.7 |
| Gemini 3.1 Pro Preview | 43.3 |
| Gemini 3.5 Flash | 43.3 |
| Grok 4.5 | 43.3 |
| Gemini 3 Flash Preview | 40.0 |
| Grok 4.1 Fast | 40.0 |
| Gemini 3.1 Flash Image Preview | 36.7 |
| Claude Fable 5 | 33.3 |
| Grok Build 0.1 | 33.3 |
| Grok 4.3 | 30.0 |
| GPT-5.6 Sol | 23.3 |
| GPT-5.6 Terra | 20.0 |
| GPT-5.4 | 13.3 |
| GPT-5.5 | 13.3 |
| Opus 4.8 | 10.0 |
| Opus 4.6 | 6.7 |
| Qwen3.6 Plus (2026-04-02) | 6.7 |
| Sonnet 5 | 3.3 |
| GLM-5.1 | 3.3 |
| GPT-5.6 Luna Pro | 3.3 |
| Haiku 4.5 | 0.0 |
| Opus 4.7 | 0.0 |
| Sonnet 4.6 | 0.0 |
| Gemini 2.5 Pro | 0.0 |
| GLM-5.2 | 0.0 |
| GLM-5 | 0.0 |
| Qwen3.6 Plus Preview | 0.0 |