Artificial Analysis · MMMU Pro
The Frontier
Best score over time · one chart, every benchmark
Full rankings
35 models tested · sorted by score
| # | Model | Score |
|---|---|---|
| 1 | 87.7 | |
| 2 | 86.9 | |
| 3 | 86.0 | |
| 4 | 85.6 | |
| 5 | 82.8 | |
| 6 | 82.4 | |
| 7 | 80.7 | |
| 8 | 80.5 | |
| 9 | 80.5 | |
| 10 | 79.7 | |
| 11 | 79.0 | |
| 12 | 78.6 | |
| 13 | 78.5 | |
| 14 | 77.0 | |
| 15 | 76.4 | |
| 16 | 76.4 | |
| 17 | 76.3 | |
| 18 | 75.0 | |
| 19 | 73.4 | |
| 20 | 73.1 | |
| 21 | 70.3 | |
| 22 | 70.1 | |
| 23 | 69.2 | |
| 24 | 66.8 | |
| 25 | 65.4 | |
| 26 | 64.9 | |
| 27 | 63.2 | |
| 28 | 62.1 | |
| 29 | 56.8 | |
| 30 | 52.9 | |
| 31 | 52.7 | |
| 32 | 43.0 | |
| 33 | 37.5 | |
| 34 | 25.8 | |
| 35 | 14.5 |
Score distribution
Where models cluster
Correlated benchmarks
Pearson r · original research
Benchmarks that track with Artificial Analysis · MMMU Pro
Pearson correlation across models scored on both benchmarks. Closer to 1 = strongly predictive.
Frequently asked
About Artificial Analysis · MMMU Pro
What does Artificial Analysis · MMMU Pro measure?
Artificial Analysis · MMMU Pro is a knowledge benchmark in the BenchGecko catalog. 35 AI models have been tested on it. Scores range from 14.5 to 87.7 out of 100.
Which model leads on Artificial Analysis · MMMU Pro?
Claude Opus 5.5 from Anthropic leads Artificial Analysis · MMMU Pro with a score of 87.7. The median score across 35 tested models is 75.0.
Is Artificial Analysis · MMMU Pro saturated?
No · the top score is 87.7 out of 100 (88%). There is still meaningful room for improvement on Artificial Analysis · MMMU Pro.
Does Artificial Analysis · MMMU Pro predict performance on other benchmarks?
Yes · Artificial Analysis · MMMU Pro scores correlate 0.99 with HLE across 5 shared models. Models that do well on Artificial Analysis · MMMU Pro tend to do well on HLE.
How often is Artificial Analysis · MMMU Pro data refreshed?
BenchGecko pulls updates daily. New model scores on Artificial Analysis · MMMU Pro appear as soon as they are published by Epoch AI or the model provider.
- Category
- Knowledge
- Max score
- 100
- Models
- 35
- Updated
- 2026-10-06
Top on Artificial Analysis · MMMU Pro
Claude Opus 5.5 · 87.7GPT-6 Astra · 86.9GPT-6.1 Sol · 86.0Gemini 3.8 Flash · 85.6Qwen3.8 Max (0902) · 82.8More knowledge benchmarks
Same category · related evaluations