Video Understanding on Video-MME Long (Accuracy w/o/w Subtitles)
67.4Accuracy (Long, wo Sub)Gemini-1.5-Pro
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Gemini-1.5-ProSize=Undisclosed2026.02 | 67.4 | 77.4 | |
| Gemini 1.5 ProAccess=Proprietary2025.12 | 67.4 | 77.4 | |
| GPT4-oSize=Undisclosed2026.02 | 65.3 | 72.1 | |
| GPT-4oAccess=Proprietary2025.12 | 65.3 | 72.1 | |
| LLaVA-OneVision (72B)Size=72B, Frames=322026.02 | 60 | 62.4 | |
| InternVL3*Size=8B, Frames=5122026.02 | 57.7 | 67.2 | |
| LongVPO (128f) Stage 2Size=8B, Frames=768, Backbone=InternVL32026.02 | 57.7 | 67.7 | |
| VideoLLaMA2 (72B)Size=72B, Frames=162026.02 | 57.6 | 59 | |
| LongVPO (128f) Stage 1Size=8B, Frames=512, Backbone=InternVL32026.02 | 57 | 67.9 | |
| LongVPO (256f) Stage 2-iter2Size=8B, Frames=512, Backbone=InternVideo2.52026.02 | 56.1 | 67.4 | |
| LongVPO (256f) Stage 1Size=8B, Frames=512, Backbone=InternVL2.52026.02 | 55.8 | 65.1 | |
| VideoChat-FlashSize=7B, Frames=5122026.02 | 55.4 | 63.3 | |
| LongVPO (128f) Stage 2Size=8B, Frames=512, Backbone=InternVL2.52026.02 | 55.3 | 64.2 | |
| LongVPO (256f) Stage 2-iter1Size=8B, Frames=512, Backbone=InternVideo2.52026.02 | 54.9 | 67.1 | |
| LongVPO (512f) Stage 1Size=8B, Frames=512, Backbone=InternVideo2.52026.02 | 54.9 | 67.1 | |
| NVILASize=7B, Frames=2562026.02 | 54.8 | 63.3 | |
| Zoom-ZeroAccess=Open-Source, Model Scale=7B-8B2025.12 | 54.8 | 64.2 | |
| LongVPO (256f) Stage 1Size=8B, Frames=512, Backbone=InternVideo2.52026.02 | 54.3 | 67 | |
| LongVPO (128f) Stage 1Size=8B, Frames=512, Backbone=InternVL2.52026.02 | 53.8 | 62.8 | |
| GPT4-VSize=Undisclosed2026.02 | 53.5 | 56.9 | |
| VideoChat-R1Access=Open-Source, Model Scale=7B-8B2025.12 | 53.4 | 62.3 | |
| TVG-R1Access=Open-Source, Model Scale=7B-8B2025.12 | 52.7 | 62.4 | |
| InternVideo2.5*Size=8B, Frames=5122026.02 | 52.6 | 65.1 | |
| MiniCPM-v2.6Size=8B, Frames=642026.02 | 51.8 | 56.3 | |
| InternVL2.5*Size=8B, Frames=5122026.02 | 51.1 | 57.2 | |
| Qwen2.5-VLAccess=Open-Source, Model Scale=7B-8B2025.12 | 51.1 | 62 | |
| mPLUG-Owl3Size=7B, Frames=162026.02 | 50.1 | — | |
| Video-XLSize=7B, Frames=2,0482026.02 | 49.2 | — | |
| Video-CCAMSize=9B, Frames=962026.02 | 46.7 | 49.9 | |
| KangarooSize=8B, Frames=642026.02 | 46.7 | 59.3 | |
| LongVASize=7B, Frames=1282026.02 | 46.2 | 47.6 | |
| LongVAAccess=Open-Source, Model Scale=7B-8B2025.12 | 46.2 | — |