Mathematical Reasoning on AIME 2025 (Accuracy, Avg Tokens)
86.67AccuracyGemma-4 31B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Gemma-4 31BEvaluation Mode=Reasoning+Rotate2Think2026.06 | 86.67 | 4,240 | |
| Gemma-4 31BEvaluation Mode=Reasoning2026.06 | 83.33 | 7,170 | |
| Qwen3 4BEvaluation Mode=Reasoning+Rotate2Think2026.06 | 80 | 20,894 | |
| Phi-4 14BEvaluation Mode=Reasoning+Rotate2Think2026.06 | 76.67 | 17,108 | |
| Phi-4 14BEvaluation Mode=Reasoning2026.06 | 73.33 | 15,053 | |
| Gemma-4 31BEvaluation Mode=Base+Rotate2Think2026.06 | 73.33 | 4,110 | |
| Qwen3 4BEvaluation Mode=Reasoning2026.06 | 70 | 20,944 | |
| Gemma-4 31BEvaluation Mode=Base2026.06 | 70 | 2,927 | |
| Gemma-4 E4BEvaluation Mode=Reasoning+Rotate2Think2026.06 | 50 | 5,891 | |
| Qwen3 4BEvaluation Mode=Base+Rotate2Think2026.06 | 46.67 | 8,299 | |
| Qwen3 4BEvaluation Mode=Base2026.06 | 43.33 | 10,444 | |
| Gemma-4 E4BEvaluation Mode=Base+Rotate2Think2026.06 | 40 | 5,408 | |
| Gemma-4 E4BEvaluation Mode=Reasoning2026.06 | 40 | 6,968 | |
| Gemma-4 E4BEvaluation Mode=Base2026.06 | 36.67 | 3,438 | |
| Phi-4 14BEvaluation Mode=Base+Rotate2Think2026.06 | 20 | 1,639 | |
| Phi-4 14BEvaluation Mode=Base2026.06 | 16.67 | 1,672 |