MusTBENCH - Global Temporal Ordering: leaderboard

MusTBench (2026) tests temporal grounding of audio-language models in full-length music tracks, zero-shot. Global temporal ordering: choose the chronological order of three transition descriptions from the same song among the six permutations. Accuracy in percent; higher is better.

Metric: Global Temporal Ordering (GTO, 198 questions): accuracy (%) at choosing the chronological order of three transition descriptions from the same song among the six permutations, 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 June 2028. 12 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Qwen 3 Omni 30B A3B63.6#847
2Gemini 3 Flash47#175
3Qwen2.5-Omni-7B46.5#1884
4Gemini 2.5 Pro46#276
5Gemini 2.5 Flash42.4#463
6Music Flamingo41.4#866
7AF-Next41.4
8GPT-Audio 1.537.9#501
9Qwen2.5-Omni-3B37.4#3519
10GPT-Audio36.9#607
11Gemini 3 Pro27.3#145
12Phi-4 Multimodal Instruct13.1#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-global-temporal-ordering · How It Works · Data refreshed daily, snapshot 2026-11-21.