List your company

Mathematical Reasoning on FormInv v1 (50 theorems, 366 items)

96.4AccuracyDeepSeek V3
85.58488.39291.294.008May 27, 2026

Evaluation results

MethodLinks
DeepSeek V3Tier=Flagship2026.0596.46.982
o4-miniTier=Reasoning2026.0596.47.680
Gemini 2.5 Flash†Tier=Flagship2026.05958.480
GPT-4o-miniTier=Efficient2026.0594.38.778
GPT-4oTier=Flagship2026.0594.2782
Claude Sonnet 4.6Tier=Flagship2026.0593.210.774
DeepSeek R1Tier=Reasoning2026.05938.878
Llama 3.3 70BTier=Open2026.0590.513.566
Claude Haiku 4.5Tier=Efficient2026.058619.450