Vision & documents

MMVU leaderboard

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap

As of 2026-10-07, the highest measured score on MMVU is 82.4% by Qwen3.8 Max. 14 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Qwen3.8 Max	82.4%	measured
2	GLM-5.3-Flash	80.5%	measured
3	Kimi K2.5	80.4%	measured
4	Kimi K3	80.0%	estimated ± 2.9 pp, medium confidence
5	dots3-note Preview	79.9%	measured
6	Seed 2.1 Pro	79.1%	estimated ± 2.9 pp, medium confidence
7	Qwen3.8-Omni-Flash	78.7%	estimated ± 2.9 pp, medium confidence
8	Qwen3.8-Flash-Next	78.1%	estimated ± 2.9 pp, medium confidence
9	Qwen3.7 Plus	77.9%	estimated ± 2.9 pp, medium confidence
10	Seed 2.1 Turbo	77.8%	estimated ± 2.9 pp, medium confidence
11	Qwen3.8-27B	77.7%	estimated ± 2.9 pp, medium confidence
12	Qwen3.5 397B	77.0%	estimated ± 2.9 pp, medium confidence
13	Qwen3.6 Plus	76.7%	estimated ± 2.9 pp, medium confidence
14	Kimi K2.6	76.4%	estimated ± 2.9 pp, medium confidence
15	Gemini 3 Pro	75.9%	estimated ± 2.9 pp, medium confidence
16	Qwen3.5-122B-A10B	74.7%	measured
17	GPT-5.2	73.9%	estimated ± 2.9 pp, low confidence
18	Qwen3.5-27B	73.3%	measured
19	Qwen3.5-35B-A3B	72.3%	measured
20	Gemma 4 12B	72.1%	estimated ± 2.9 pp, low confidence
21	Claude Opus 4.5	68.8%	estimated ± 2.9 pp, low confidence

MMVU leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General