# MMMU / MMMU-Pro

College exam-style questions that come with pictures, such as charts, diagrams, maps and sheet music, across 30 subjects. It checks whether an AI can understand images and reason about them like a university student.

Category: multimodal. Type: leaderboard. Status: active.
Metric: MMMU-Pro overall accuracy: share of questions answered correctly (%), on the harder version with ten answer options and questions embedded in images (higher is better).
Source: MMMU team (mmmu-benchmark.github.io) — https://mmmu-benchmark.github.io/leaderboard_data.json, results as of 2026-11-10, retrieved 2026-11-17, license: not stated.

| Rank | Model | Type | Date | MMMU-Pro Overall | MMMU Val Overall |
|---|---|---|---|---|---|
| 1 | Chance Vision 1.5 | proprietary | 2026-07-01 | 86.9 | - |
| 2 | GPT-5.4 Thinking w/ tools | proprietary | 2026-03-05 | 82.1 | - |
| 3 | GPT-5.4 Thinking w/o tools | proprietary | 2026-03-05 | 81.2 | - |
| 4 | Gemini 3.0 Pro | proprietary | 2025-11-18 | 81.0 | - |
| 5 | Gemini 3.1 Pro Thinking (High) | proprietary | 2026-02-19 | 80.5 | - |
| 6 | Muse Spark Thinking | proprietary | 2026-04-08 | 80.4 | - |
| 7 | GPT-5.2 Thinking w/o Python | proprietary | 2025-12-11 | 80.4 | - |
| 8 | GPT-5.2 Thinking w/o tools | proprietary | 2025-12-11 | 79.5 | - |
| 9 | GPT-5.1 Thinking | proprietary | 2025-11-13 | 79.0 | - |
| 10 | GPT-5 w/ thinking | proprietary | 2025-08-07 | 78.4 | 84.2 |
| 11 | Claude Opus 4.6 w/ tools | proprietary | 2026-02-05 | 77.3 | - |
| 12 | Gemma 4 31B | open source | 2026-04-02 | 76.9 | - |
| 13 | o3 | proprietary | 2025-04-16 | 76.4 | 82.9 |
| 14 | GPT-5.1 | proprietary | 2025-11-13 | 76.0 | 85.4 |
| 15 | Claude Sonnet 4.6 w/ tools | proprietary | 2026-02-17 | 75.6 | - |
| 16 | Claude Sonnet 4.6 w/o tools | proprietary | 2026-02-17 | 74.5 | - |
| 17 | Claude Opus 4.6 w/o tools | proprietary | 2026-02-05 | 73.9 | - |
| 18 | Claude Opus 4.5 | proprietary | 2025-11-24 | 73.9 | 80.7 |
| 19 | Gemma 4 26B A4B | open source | 2026-04-02 | 73.8 | - |
| 20 | dots.vlm1 | open source | 2025-08-05 | 70.1 | 80.1 |
| 21 | Claude Sonnet 4.5 | proprietary | 2025-09-29 | 68.9 | 77.8 |
| 22 | Qwen3-VL 235B-A22B | open source | 2025-09-22 | 68.1 | 78.7 |
| 23 | Gemini 2.5 Pro 05-06 | proprietary | 2025-05-06 | 68.0 | 79.6 |
| 24 | Seed 1.5-VL Thinking | proprietary | 2025-05-13 | 67.6 | 77.9 |
| 25 | Seed 1.6-Thinking | proprietary | 2025-06-25 | 66.4 | 74.8 |

Notes: Rows are models with an MMMU-Pro score, sorted by MMMU-Pro overall (the page itself lists entries unsorted by default); rank is computed from that order. Human-expert and random baselines excluded (human expert high: 85.4 on MMMU-Pro). Nearly all scores are reported by model authors ('source: author'), not independently verified. '-' means no MMMU validation score listed. as_of is the latest model date in the data file, which the page uses as its 'Last updated' value.
