MuseBench: Benchmarking Intent-Level Audiovisual Arts Understanding in MLLMs: leaderboard
MuseBench evaluates multimodal large language models on audiovisual arts understanding, using video-based questions across cinematic, visual, stage, and game arts. The pinned metric is Overall ACC (%), measuring answer accuracy on the benchmark’s single-select and variable-option multi-select items in the mirrored source tables. Source: BenchmarkList’s attributed result table.
Metric: Overall ACC (%). Source: benchmarklist.com. Saturation forecast: Around 2030. 28 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.6 | 48.29 | #67 |
| 2 | Qwen 3.5 Plus | 47.27 | #133 |
| 3 | Qwen 3.5 397B A17B | 44.76 | #154 |
| 4 | GPT-5.4 | 44.58 | #83 |
| 5 | InternVL3-78B | 37.81 | #361 |
| 6 | Gemini 3.1 Pro (Preview) | 36.89 | #64 |
| 7 | InternVL3-8B | 33.07 | #624 |
| 8 | Qwen2.5-Omni-7B | 32.7 | #633 |
| 9 | Gemma 4 E4B | 27.61 | #762 |
| 10 | Grok 4.1 | 20.54 | #239 |
| 11 | Kimi K2.5 | 19.91 | #152 |
| 12 | GLM-4.5V | 17.13 | #358 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=musebench-benchmarking-intent-level-audiovisual-arts-understanding-in-mllms · How It Works · Data refreshed daily, snapshot 2026-11-05.