VideoOdyssey-AV - Action Recognition: leaderboard

VideoOdyssey (2026) tests ultra-long video understanding with human-written multiple-choice questions on videos averaging 109 minutes, each annotated with how long a viewer must watch continuously to answer. This board is accuracy in percent on the audio-visual subset questions about recognizing actions; higher is better.

Metric: Multiple-choice accuracy (%) on the 60 action recognition questions (identify the actions of humans or other objects in the video, based on both visual and audio cues) of VideoOdyssey-AV (human-written questions on 100 videos of 60 to 255 minutes that need both visual and audio cues; 1,062 in all), omni-modal models given video and audio, 64 frames (VITA-1.5-7B 16); higher is better. Source: arxiv.org. Saturation forecast: Around September 2027. 12 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Flash45#175
2Gemini 2.5 Pro40#276
3Gemini 3.1 Pro (Preview)40#99
4Qwen 3.5 Omni Plus38.3#254
5Qwen 3 Omni 30B A3B36.7#847
6VITA-1.5-7B30#1853
7Unified-IO-2 XL28.3
8Ola-7B23.3#1557
9Unified-IO-2 XXL23.3
10Unified-IO-2 L21.7
11OneLLM21.4#2372
12VideoLLaMA2-7B16.6#2385

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=videoodyssey-av-action-recognition · How It Works · Data refreshed daily, snapshot 2026-11-21.