List your company

Mathematical Reasoning on AIME 2026 (Accuracy, Average tokens generated)

86.67AccuracyGemma-4 31B
6.933227.634148.33569.0359Jun 2, 2026

Evaluation results

MethodLinks
Gemma-4 31BEvaluation Mode=Reasoning+Rotate2Think2026.0686.673,098
Gemma-4 31BEvaluation Mode=Base+Rotate2Think2026.0683.332,036
Gemma-4 31BEvaluation Mode=Reasoning2026.0683.337,894
Phi-4 14BEvaluation Mode=Reasoning+Rotate2Think2026.068017,540
Gemma-4 31BEvaluation Mode=Base2026.06802,119
Phi-4 14BEvaluation Mode=Reasoning2026.0676.6717,077
Qwen3 4BEvaluation Mode=Reasoning+Rotate2Think2026.0673.3321,076
Qwen3 4BEvaluation Mode=Reasoning2026.067020,609
Qwen3 4BEvaluation Mode=Base+Rotate2Think2026.0653.338,855
Gemma-4 E4BEvaluation Mode=Reasoning+Rotate2Think2026.0653.335,586
Gemma-4 E4BEvaluation Mode=Base+Rotate2Think2026.0646.675,422
Gemma-4 E4BEvaluation Mode=Reasoning2026.0643.336,829
Qwen3 4BEvaluation Mode=Base2026.06409,359
Gemma-4 E4BEvaluation Mode=Base2026.0633.333,691
Phi-4 14BEvaluation Mode=Base+Rotate2Think2026.0616.671,515
Phi-4 14BEvaluation Mode=Base2026.06101,380