Video Understanding on WorldSense
52.01ScoreQwen3-Omni-30B-A3B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Qwen3-Omni-30B-A3BSize=30B2026.02 | 52.01 | — | |
| OmniVinciSize=7B2026.02 | 48.23 | — | |
| Qwen2.5-OmniSize=7B2026.02 | 45.4 | — | |
| OmniFysicsSize=3B2026.02 | 45.39 | — | |
| Qwen2.5-OmniSize=3B2026.02 | 44.45 | — | |
| VISIONCOACH-7BModel Category=Tool-free, Model Scale=7B2026.03 | 43.8 | 41.8 | |
| GPT-4oModel Category=Proprietary2026.03 | 42.6 | — | |
| GPT-4o2026.03 | 42.6 | — | |
| GPT-4o2026.06 | 42.6 | — | |
| SER-7BParameters=7B2026.06 | 42.3 | 42.4 | |
| EgoR1Model Category=Tool-calling, RAG=false2026.03 | 42 | 40.3 | |
| Open-o3 + MCoT2026.03 | 41.5 | 40.8 | |
| InternVL-2.5-8BParameters=8B2026.06 | 39.6 | 38.5 | |
| Motion-o2026.03 | 39.2 | 38.5 | |
| Open-o3 + MCoTgrounding reward=false2026.03 | 38.9 | 38.2 | |
| VideoRFT-7BModel Category=Tool-free, Model Scale=7B2026.03 | 38.2 | 36.6 | |
| VideoRFT-7B2026.03 | 38.2 | 36.6 | |
| VideoRFT-7BParameters=7B2026.06 | 38.2 | 36.6 | |
| Motion-ogrounding reward=false2026.03 | 37.8 | 37.1 | |
| Open-o3-video-7BModel Category=Tool-free, Model Scale=7B2026.03 | 37.5 | 36.8 | |
| Open-o3 Video2026.03 | 37.5 | 36.8 | |
| Open-o3-Video2026.06 | 37.5 | 36.8 | |
| VideoLLaMA3-7BParameters=7B2026.06 | 37.3 | 38.1 | |
| Qwen2.5-VL-7BModel Category=Tool-free, Model Scale=7B2026.03 | 36.1 | 33.7 | |
| Qwen2.5-VL-7Bvariant=base2026.03 | 36.1 | 33.7 | |
| Qwen2.5-VL-7BParameters=7B, Model Version=Base model2026.06 | 36.1 | 33.7 | |
| VideoR1-7BModel Category=Tool-free, Model Scale=7B2026.03 | 35.5 | 32.8 | |
| VideoR1-7B2026.03 | 35.5 | 32.8 | |
| VideoR1-7BParameters=7B2026.06 | 35.5 | 32.8 | |
| Unified-IO-2 XXLSize=7B2026.02 | 25.9 | — | |
| Unified-IO-2 XLSize=3B2026.02 | 24.7 | — | |
| Unified-IO-2 LSize=1B2026.02 | 23.3 | — | |
| LongVT-RL-7BModel Category=Tool-calling, Model Scale=7B2026.03 | 22.1 | 22.5 |