MMAU-Pro: leaderboard

Audio-language models answer human-written questions about real-world speech, sound and music clips of up to 10 minutes; the score is the paper's weighted average accuracy over all 5,305 instances, 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: Estimated already saturated. 18 models tracked. Updated 17 Nov 2026.

Top models

#ModelScoreOverall rank
1Qwen2.5-Omni-7B52.2#1883
2Audio Flamingo 351.7
3Ming-Lite-Omni-1.547.4#2348
4Qwen2.5-Omni-3B46.1#3596
5Audio Flamingo 242.6
6DeSTA2.5-Audio40.6#2076
7Gemma 3n E4B (IT)39.7#2432
8SALMONN-13B39.6
9Audio Reasoner39.5#3028
10Phi-4 Multimodal Instruct38.7#4209
11DeSTA2-8B-beta36.7
12Gemma 3n E2B (IT)35.4#4283
13SALMONN-7B34.5
14R1-AQA34.1#5707
15Baichuan-Omni-1.533.9#3064

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 · How It Works · Data refreshed daily, snapshot 2026-11-17.