MMAU-Pro - Speech: 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 891 speech 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: Estimated already saturated. 18 models tracked. Updated 17 Nov 2026.

Top models

#ModelScoreOverall rank
1Audio Flamingo 358.8
2Qwen2.5-Omni-7B57.4#1883
3Qwen2.5-Omni-3B53.9#3596
4DeSTA2.5-Audio49.9#2076
5Ming-Lite-Omni-1.549.1#2348
6Phi-4 Multimodal Instruct47.6#4209
7DeSTA2-8B-beta46.5
8Gemma 3n E4B (IT)44.9#2432
9Audio Reasoner44#3028
10Audio Flamingo 243
11Gemma 3n E2B (IT)41.3#4283
12SALMONN-7B38.3
13SALMONN-13B37.3
14Baichuan-Omni-1.536.5#3064
15R1-AQA33.7#5707

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