AudioBench - MMAU Mini: leaderboard
AudioBench MMAU Mini slice measuring multimodal audio understanding across speech, sound, and music tasks.
Metric: Mean Score (%). Source: github.com. Saturation forecast: Estimated already saturated. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Phi-4 Multimodal Instruct | 56.95 | #906 |
| 2 | SeaLLMs-Audio-7B | 55.85 | #1018 |
| 3 | Qwen2-Audio-7B-Instruct | 52.1 | #880 |
| 4 | Qwen-Audio-Chat | 46.05 | #1244 |
| 5 | Gemini 1.5 Flash | 44.8 | #618 |
| 6 | SALMONN-7B | 44.45 | #1293 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=audiobench-mmau-mini · How It Works · Data refreshed daily, snapshot 2026-10-20.