Mathematics Reasoning on AIME 2026 (Accuracy)
97.1Accuracy (AIME 2026)VibeThinker-3B
Evaluation Results
| Method | Links | |
|---|---|---|
| VibeThinker-3BParams=3B, CLR reasoning enhancement=true2026.06 | 97.1 | |
| VibeThinker-3BParams=3B2026.06 | 94.3 | |
| GPT-OSS-20B (high)Params=20B2026.06 | 90.2 | |
| Nemotron-3-NanoParams=30B2026.06 | 90.1 | |
| Ministral-3-Reasoning-2512Params=14B2026.06 | 85 | |
| Qwen3.5-4BParams=4B2026.06 | 84 | |
| OpenReasoning-NemotronParams=7B2026.06 | 80.2 | |
| Qwen3-4B-Thinking-2507Params=4B2026.06 | 79 | |
| Gemma-4-itParams=12B2026.06 | 77.5 | |
| Mimo7B-RL-0530Params=7B2026.06 | 76 | |
| Phi4-Reasoning-PlusParams=14B2026.06 | 73.6 | |
| Olmo-3-ThinkParams=7B2026.06 | 69.1 | |
| Hunyuan-4B-InstructParams=4B2026.06 | 57.7 | |
| SmolLM3Params=3B2026.06 | 41 |