Mathematics Reasoning on AIME 2025 (Accuracy)
96.7AIME 2025 AccuracyVibeThinker-3B
Evaluation results
| Method | Links | |
|---|---|---|
| VibeThinker-3BParams=3B, CLR reasoning enhancement=true2026.06 | 96.7 | |
| Qwen3-235B-A22B-ThinkingParams=235B2026.06 | 92.3 | |
| GPT-OSS-20B (high)Params=20B2026.06 | 91.7 | |
| VibeThinker-3BParams=3B2026.06 | 91.4 | |
| LongCat FlashParams=560B2026.06 | 90.6 | |
| Nemotron-3-NanoParams=30B2026.06 | 89.1 | |
| GPT-5 Nano (high)Params=N/A2026.06 | 85.2 | |
| GLM-4.5-AirParams=106B2026.06 | 83.3 | |
| Ministral-3-Reasoning-2512Params=14B2026.06 | 82.9 | |
| Qwen3-4B-Thinking-2507Params=4B2026.06 | 81.3 | |
| Qwen3.5-4BParams=4B2026.06 | 79.8 | |
| OpenReasoning-NemotronParams=7B2026.06 | 78.2 | |
| Gemma-4-itParams=12B2026.06 | 72.9 | |
| Mimo7B-RL-0530Params=7B2026.06 | 70.2 | |
| Phi4-Reasoning-PlusParams=14B2026.06 | 68.4 | |
| Olmo-3-ThinkParams=7B2026.06 | 67.9 | |
| Hunyuan-4B-InstructParams=4B2026.06 | 66.5 | |
| SmolLM3Params=3B2026.06 | 36.7 |