Mathematics on AIME 2024 (Score %)
90.94AIME 2024 Score (%)STEP3-VL-10B
Evaluation Results
| Method | Links | |
|---|---|---|
| STEP3-VL-10BNumber of Parameters=10B2026.01 | 90.94 | |
| InternVL 3.5Number of Parameters=8B2026.01 | 78.18 | |
| MiMo-VL RL-2508Number of Parameters=7B2026.01 | 75.36 | |
| Qwen3-VL ThinkingNumber of Parameters=8B2026.01 | 74.06 | |
| GLM-4.6V FlashNumber of Parameters=9B2026.01 | 37.92 | |
| ePFBackbone=Qwen2.5-7B-In, Weighting scheme=w2025.10 | 26.23 | |
| PFBackbone=Qwen2.5-7B-In, Weighting scheme=w2025.10 | 26.06 | |
| Best-of-NBackbone=Qwen2.5-7B-In, Weighting scheme=w2025.10 | 23.09 | |
| PFBackbone=Qwen2.5-7B-In, Weighting scheme=u2025.10 | 21.6 | |
| ePFBackbone=Qwen2.5-7B-In, Weighting scheme=u2025.10 | 21 | |
| Best-of-NBackbone=Qwen2.5-7B-In, Weighting scheme=u2025.10 | 20.2 | |
| PFBackbone=Qwen2.5-7B-In, Weighting scheme=iw2025.10 | 18.13 | |
| Beam-SearchBackbone=Qwen2.5-7B-In, Weighting scheme=w2025.10 | 17.93 | |
| Best-of-NBackbone=Qwen2.5-7B-In, Weighting scheme=iw2025.10 | 17.48 | |
| ePFBackbone=Qwen2.5-1.5B-In, Weighting scheme=w2025.10 | 17.06 | |
| ePFBackbone=Qwen2.5-7B-In, Weighting scheme=iw2025.10 | 16.06 | |
| Beam-SearchBackbone=Qwen2.5-7B-In, Weighting scheme=u2025.10 | 14 | |
| Beam-SearchBackbone=Qwen2.5-7B-In, Weighting scheme=iw2025.10 | 11.26 | |
| ePFBackbone=Qwen2.5-1.5B-In, Weighting scheme=u2025.10 | 11.2 | |
| PFBackbone=Qwen2.5-1.5B-In, Weighting scheme=w2025.10 | 11.16 | |
| Beam-SearchBackbone=Qwen2.5-1.5B-In, Weighting scheme=w2025.10 | 10.29 | |
| Base SamplingBackbone=Qwen2.5-7B-In, Weighting scheme=u2025.10 | 10 | |
| Best-of-NBackbone=Qwen2.5-1.5B-In, Weighting scheme=w2025.10 | 9.9 | |
| PFBackbone=Qwen2.5-1.5B-In, Weighting scheme=u2025.10 | 9 | |
| Beam-SearchBackbone=Qwen2.5-1.5B-In, Weighting scheme=u2025.10 | 8.4 | |
| Best-of-NBackbone=Qwen2.5-1.5B-In, Weighting scheme=u2025.10 | 7.2 | |
| PFBackbone=Qwen2.5-1.5B-In, Weighting scheme=iw2025.10 | 6.99 | |
| Beam-SearchBackbone=Qwen2.5-1.5B-In, Weighting scheme=iw2025.10 | 5.55 | |
| ePFBackbone=Qwen2.5-1.5B-In, Weighting scheme=iw2025.10 | 5.55 | |
| Best-of-NBackbone=Qwen2.5-1.5B-In, Weighting scheme=iw2025.10 | 4 | |
| Base SamplingBackbone=Qwen2.5-1.5B-In, Weighting scheme=u2025.10 | 3.33 |