MMAU-Pro - Open-ended QA: 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 625 open-ended questions, rated 1-5 by Qwen2.5-7B-Instruct and converted to percent, 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

#ModelScoreOverall rank
1Qwen2.5-Omni-7B52.3#1883
2Qwen2.5-Omni-3B47.6#3596
3Audio Flamingo 344.2
4Audio Flamingo 243.2
5Ming-Lite-Omni-1.542.7#2348
6Phi-4 Multimodal Instruct42.5#4209
7Baichuan-Omni-1.539.7#3064
8Audio Reasoner38.6#3028
9R1-AQA38.5#5707
10DeSTA2.5-Audio36.4#2076
11SALMONN-13B33.6
12SALMONN-7B31.2
13Gemma 3n E4B (IT)28.5#2432
14DeSTA2-8B-beta25.4
15BAT24.6

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