AVI-Bench - Audio-Visual Language Grounding: leaderboard

Metric: Task score (%) on audio-visual language grounding (503 videos): mean over videos of the per-frame IoU of the box predicted for the referred sounding object, 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

#ModelScore
1Gemini 2.5 Pro (Preview 05-06)35.08
2Gemini 2.5 Flash (Preview 04-17)32.79
3Gemini 1.5 Flash30.29
4Gemini 1.5 Pro29.54
5Gemini 2.0 Flash27.61
6Qwen2.5-Omni-7B8.74
7Phi-4 Multimodal Instruct2.27

Interactive version: theaggregate.ai/benchmark?slug=avi-bench-audio-visual-language-grounding · How It Works · Data refreshed daily, snapshot 2026-09-29.