Math on AIME (Accuracy)
73.4AccuracyQwen3.5-9B
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen3.5-9BParameters=9B, Variant=Thinking2026.05 | 73.4 | |
| Qwen3.5-4BParameters=4B, Variant=Thinking2026.05 | 68.3 | |
| OLMo-3-7BParameters=7B, Variant=Thinking2026.05 | 61.7 | |
| Mellum 2 (RL)Post-training Stage=RL, Parameters=2.5B/12B, Variant=Thinking2026.05 | 58.4 | |
| Ministral-3-14BParameters=14B, Variant=Thinking2026.05 | 38.3 | |
| Mellum 2 (SFT)Post-training Stage=SFT, Parameters=2.5B/12B, Variant=Thinking2026.05 | 20 |