MMAU-Pro - Spatial Audio: 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 325 binaural spatial-audio questions, 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
1Audio Flamingo 244.1
2Qwen2.5-Omni-7B41.2#1883
3Phi-4 Multimodal Instruct39.7#4209
4DeSTA2-8B-beta32.6
5Ming-Lite-Omni-1.531.7#2348
6SALMONN-13B30.8
7Qwen2.5-Omni-3B28.9#3596
8DeSTA2.5-Audio28#2076
9Audio Flamingo 326.8
10SALMONN-7B26.5
11Mellow23.7
12BAT23.7
13R1-AQA23.6#5707
14Gemma 3n E4B (IT)21.8#2432
15Baichuan-Omni-1.521.2#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-spatial-audio · How It Works · Data refreshed daily, snapshot 2026-11-17.