Long-video sustained reasoning on LVOmniBench
42.9AccuracyGemini-2.0-Flash
Evaluation results
| Method | Links | |
|---|---|---|
| Gemini-2.0-FlashModel Accessibility=Proprietary2026.05 | 42.9 | |
| LatentOmniModel Accessibility=Open-source, Training/Reasoning Strategy=Latent reasoning2026.05 | 35.1 | |
| MiniCPM-o-7BModel Accessibility=Open-source2026.05 | 34.8 | |
| Qwen2.5-Omni-7B + Vanilla SFTModel Accessibility=Open-source, Training/Reasoning Strategy=Vanilla SFT2026.05 | 33.2 | |
| Baichuan-Omni-1.5Model Accessibility=Open-source2026.05 | 32.8 | |
| HumanOmniV2-7BModel Accessibility=Open-source2026.05 | 32.3 | |
| Qwen2.5-Omni-7B + Explicit Text CoTModel Accessibility=Open-source, Training/Reasoning Strategy=Explicit Text CoT2026.05 | 32.1 | |
| Qwen2.5-Omni-7BModel Accessibility=Open-source, Training/Reasoning Strategy=Base2026.05 | 32 | |
| VideoLLaMA2-7BModel Accessibility=Open-source2026.05 | 27 |