MMAU-Pro - Instruction Following: leaderboard
Audio-language models answer human-written questions about real-world speech, sound and music clips of up to 10 minutes; the score is accuracy on the 87 spoken format-constraint instructions, checked by regex, from the authors' runs. Unlike the original MMAU, clips are long, can be mixed or multiple, and include spatial audio, voice chat and open-ended questions.
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Around February 2027. 18 models tracked. Updated 17 Nov 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Phi-4 Multimodal Instruct | 65.4 | #4209 |
| 2 | Qwen2.5-Omni-7B | 61.3 | #1883 |
| 3 | Qwen2.5-Omni-3B | 58.4 | #3596 |
| 4 | Ming-Lite-Omni-1.5 | 48.2 | #2348 |
| 5 | Baichuan-Omni-1.5 | 47.2 | #3064 |
| 6 | DeSTA2.5-Audio | 46.5 | #2076 |
| 7 | R1-AQA | 44.2 | #5707 |
| 8 | Audio Reasoner | 43.4 | #3028 |
| 9 | DeSTA2-8B-beta | 41.5 | |
| 10 | SALMONN-13B | 38.5 | |
| 11 | Gemma 3n E4B (IT) | 36.4 | #2432 |
| 12 | SALMONN-7B | 33.9 | |
| 13 | Audio Flamingo 3 | 33.3 | |
| 14 | BAT | 31.8 | |
| 15 | GAMA | 31.7 | #8292 |
No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.
Interactive version: theaggregate.ai/benchmark?slug=mmau-pro-instruction-following · How It Works · Data refreshed daily, snapshot 2026-11-17.