VidOmni-Bench (Video + Audio): leaderboard
VidOmni-Bench (Kim et al., 2026) run with the video and its audio track as input: the Video-LLM verifies, sentence by sentence, whether each event in a dense caption of the video is supported, with human-labelled incorrect sentences as hard negatives. F1 in percent for detecting the incorrect sentences, averaged over video-caption pairs; higher is better.
Metric: F1 (%; sentence-wise event verification: for each of the dense captions written by five Video-LLMs for the 500 videos (4 s to 90 min, five complexity types), the model labels every sentence Correct or Incorrect and the human-labelled Incorrect sentences are the detection target; precision, recall and F1 are computed per video-caption pair and averaged over pairs; video with its audio track). Source: arxiv.org. Saturation forecast: Around March 2027. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash | 41.7 | #159 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=vidomni-bench-video-plus-audio · How It Works · Data refreshed daily, snapshot 2026-11-12.