Long-form Audio-Video Understanding on OmniVideoBench up to 30min
26.2Latency (10, 30] minMiniCPM-o 2.6
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MiniCPM-o 2.6LLM Backbone Scale=7-8B2026.06 | 26.2 | 29.7 | |
| Qwen2.5-Omni-7BLLM Backbone Scale=7-8B2026.06 | 26.7 | 29.3 | |
| VideoLLaMA2-7BLLM Backbone Scale=7-8B2026.06 | 28.3 | 29.2 | |
| HumanOmni-7BLLM Backbone Scale=7-8B2026.06 | 29.3 | 30.5 | |
| MiniCPM-o 4.5LLM Backbone Scale=7-8B2026.06 | 31.6 | 36.9 | |
| OmniZipLLM Backbone Scale=7-8B2026.06 | 32.3 | 36.1 | |
| Baichuan-Omni-1.5LLM Backbone Scale=7-8B2026.06 | 32.4 | 30.7 | |
| video-SALMONN 2+ 7BLLM Backbone Scale=7-8B2026.06 | 34.6 | 37.4 | |
| AVOCLLM Backbone Scale=7-8B, Global token budget (K)=10240, Modality token ratio (Kvideo:Kaudio)=2:12026.06 | 39.8 | 42.3 |