Atlas

Benchmarks

← All benchmarks

ChessBench — Pin, Historical Output Repair

Games · 2026-03-25

The 30-puzzle Pin 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)

ModelScore
Gemini 3.7 Flash63.3
Gemini 3.6 Flash60.0
GPT-6 Astra50.0
Grok 4.2046.7
Gemini 3.1 Pro Preview43.3
Gemini 3.5 Flash43.3
Grok 4.543.3
Grok 4.1 Fast40.0
Gemini 3.1 Flash Image Preview36.7
Claude Fable 533.3
Grok Build 0.133.3
Grok 4.330.0
GPT-5.6 Sol23.3
GPT-5.6 Terra20.0
Claude Opus 516.7
GPT-5.413.3
GPT-5.513.3
Opus 4.810.0
Opus 4.66.7
Qwen3.6 Plus (2026-04-02)6.7
Sonnet 53.3
GLM-5.13.3
GPT-5.6 Luna3.3
Opus 4.70.0
Sonnet 4.60.0
Gemini 2.5 Pro0.0
GLM-5.20.0
GLM-50.0
Qwen3.6 Plus Preview0.0
Loading Atlas data…