MusTBENCH - Local Transition Recognition: leaderboard

MusTBench (2026) tests temporal grounding of audio-language models in full-length music tracks, zero-shot. Local transition recognition: pick the description of the musical transition at a given timestamp, with distractors from other transitions of the same song. Accuracy in percent; higher is better.

Metric: Local Transition Recognition (LTR, 208 questions): multiple-choice accuracy (%) at picking the description of the musical transition at a given timestamp, with distractors from other transitions of the same song, on the MusTBench test set (full-length MTG-Jamendo and Slakh2100 tracks, music-expert-validated questions), zero-shot with the full track and a task instruction, temperature 0, unparseable timestamps and intervals counted wrong; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 12 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 2.5 Pro62.5#276
2Gemini 2.5 Flash56.7#463
3Music Flamingo56.3#866
4Gemini 3 Pro54.3#145
5Qwen 3 Omni 30B A3B53.4#847
6Gemini 3 Flash51.4#175
7GPT-Audio 1.551#501
8Qwen2.5-Omni-7B45.7#1884
9AF-Next44.7
10GPT-Audio42.3#607
11Qwen2.5-Omni-3B32.7#3519
12Phi-4 Multimodal Instruct28.4#4154

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=mustbench-local-transition-recognition · How It Works · Data refreshed daily, snapshot 2026-11-21.