Benchmark
OmniVideoBench
What is OmniVideoBench?
1,000 audio-visual reasoning questions over 628 videos (seconds to 30 minutes long), targeting synergistic audio-visual understanding in Omni MLLMs, by the…
- Released
- 2025-10-12
- Evaluates
- multimodal
- Openness
- unknown
- Importer
- Model reports
- Reported scores
- 19
Source provenance
- Original evidence https://arxiv.org/abs/2510.10689
OmniVideoBench paper
OmniVideoBench results and reported scores
Model reports
| Model | Organization | Reported value | Reported | Evidence |
|---|---|---|---|---|
| Baichuan-Omni-1.5 | Baichuan AI | 30.7 | 2026-03-05 | evidence |
| DeepSeek-V3.1 | DeepSeek | 27.6 | 2026-03-05 | evidence |
| Gemini-2.0-Flash | 41.5 | 2026-03-05 | evidence | |
| Gemini-2.0-Flash | 31.3 | 2026-03-05 | evidence | |
| Gemini-2.5-Flash | 52.4 | 2026-03-05 | evidence | |
| Gemini-2.5-Pro | 58.9 | 2026-03-05 | evidence | |
| Gemini-3.0-Flash | 55.1 | 2026-03-05 | evidence | |
| Gemini-3.0-Pro | 61.8 | 2026-03-05 | evidence | |
| HumanOmni-7B | HumanOmni | 30.5 | 2026-03-05 | evidence |
| MiniCPM-o | OpenBMB | 29.7 | 2026-03-05 | evidence |
| OmniVinci-9B | OmniVinci | 32.1 | 2026-03-05 | evidence |
| Qwen2.5-Omni-7B | Alibaba | 29.3 | 2026-03-05 | evidence |
| Qwen2.5-Omni-7B | Alibaba | 26.4 | 2026-03-05 | evidence |
| Qwen2.5-VL-32B | Alibaba | 31.8 | 2026-03-05 | evidence |
| Qwen2.5-VL-72B | Alibaba | 29.5 | 2026-03-05 | evidence |
| Qwen2.5-VL-7B | Alibaba | 29.8 | 2026-03-05 | evidence |
| Qwen3-Omni-30B-A3B | Alibaba | 38.4 | 2026-03-05 | evidence |
| VideoLLaMA2-7B | Alibaba | 29.2 | 2026-03-05 | evidence |
| VITA-1.5-7B | VITA | 30.5 | 2026-03-05 | evidence |
Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.
Which sources cite OmniVideoBench?
- OmniVideoBench v3 results table NJU-LINK (Nanjing University)