AVI-Bench: leaderboard
AVI-Bench (2026) tests audio-visual intelligence in omni-modal LLMs over 5,864 samples and 14 tasks arranged by cognitive stage: perception, understanding, reasoning and primitive sensation of unfamiliar, low-semantic stimuli. The score is the mean of the four stage averages, in percent; higher is better.
Metric: Overall score (%): mean of the four stage averages (perception, understanding, reasoning, primitive sensation), on AVI-Bench (5,864 audio-visual items in 14 tasks over four cognitive stages), zero-shot with the same prompts for every omni-modal model; task scores normalized to percentages; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 26 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro (Preview 05-06) | 57.21 | #275 |
| 2 | Gemini 2.5 Flash (Preview 04-17) | 46.02 | #450 |
| 3 | Gemini 2.0 Flash | 44.97 | #704 |
| 4 | Qwen2.5-Omni-7B | 41.33 | #1789 |
| 5 | Gemini 1.5 Pro | 40.38 | #1049 |
| 6 | Gemini 1.5 Flash | 38.42 | #1668 |
| 7 | Phi-4 Multimodal Instruct | 29.04 | #3882 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=avi-bench · How It Works · Data refreshed daily, snapshot 2026-11-13.