AVI-Bench - Visual-Referenced Audio Retrieval: leaderboard
Metric: Task score (%) on visual-referenced audio retrieval (264 queries over 10 candidates): retrieval score scaled by a repetition penalty and by a confidence factor for oversized predicted sets, 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 |
|---|---|---|
| 1 | Gemini 2.5 Pro (Preview 05-06) | 76.04 |
| 2 | Gemini 2.0 Flash | 39.26 |
| 3 | Gemini 2.5 Flash (Preview 04-17) | 38.83 |
| 4 | Qwen2.5-Omni-7B | 35.49 |
| 5 | Gemini 1.5 Pro | 25.17 |
| 6 | Gemini 1.5 Flash | 16.8 |
| 7 | Phi-4 Multimodal Instruct | 3.69 |
Interactive version: theaggregate.ai/benchmark?slug=avi-bench-visual-referenced-audio-retrieval · How It Works · Data refreshed daily, snapshot 2026-09-29.