General Multi-task Video Understanding on VideoMME w/o sub
77.4Average AccuracyGemini1.5 Pro
Evaluation results
| Method | Links | |
|---|---|---|
| Gemini1.5 ProSize=-2026.04 | 77.4 | |
| G2F-RAGSize=8B, Base Model=InternVL3.52026.04 | 72 | |
| Qwen3-VLSize=8B2026.04 | 71.4 | |
| G2F-RAGSize=7B, Base Model=Qwen2.5-VL2026.04 | 70.6 | |
| G2F-RAGSize=14B, Base Model=InternVL3.52026.04 | 70.6 | |
| InternVL3Size=14B2026.04 | 70.4 | |
| G2F-RAGSize=4B, Base Model=InternVL3.52026.04 | 70.1 | |
| Qwen3-VLSize=4B2026.04 | 69.3 | |
| GPT-4o (2024-05-13)Size=-2026.04 | 67.9 | |
| MiniCPM-V 4.5Size=8B2026.04 | 67.9 | |
| InternVL3.5Size=14B2026.04 | 67.9 | |
| InternVL3Size=8B2026.04 | 66 | |
| InternVL3.5Size=8B2026.04 | 66 | |
| InternVL3.5Size=4B2026.04 | 65.4 | |
| InternVideo2.5Size=7B2026.04 | 65.1 | |
| Qwen2.5-VLSize=7B2026.04 | 65.1 | |
| G2F-RAGSize=7B, Base Model=LLaVA-Video2026.04 | 64.5 | |
| InternVL2.5Size=8B2026.04 | 63.7 | |
| LLaVA-VideoSize=7B2026.04 | 63.7 | |
| Qwen2.5-VLSize=3B2026.04 | 61.5 | |
| VideoLLaMA 3Size=7B2026.04 | 61 | |
| MiniCPM-V 2.6Size=8B2026.04 | 59.7 |