LLM AtlasLLM AtlasSearch models

Benchmark

MMMU

vision

Massive Multi-discipline Multimodal Understanding. Tests visual reasoning across 30+ subjects with images, charts, and diagrams.

Interpretation

MMMU is a vision benchmark evaluating visual understanding and analysis capabilities. It ranks 11 models from GPT-5.4 (78) to Llama 4 Maverick (65). This benchmark contributes to the vision scoring on model pages and rankings.

Methodology: 11,500 questions across 30 subjects requiring college-level multimodal reasoning with images, diagrams, and charts.

Source: https://mmmu-benchmark.github.io/