MMMU-Pro
A more robust multi-discipline multimodal understanding benchmark that enhances MMMU through a three-step process: filtering text-only answerable questions, augmenting candidate options, and introducing vision-only input settings. Achieves significantly lower model performance (16.8-26.9%) compared to original MMMU, providing more rigorous evaluation that closely mimics real-world scenarios.
Gemini 3.5 Flash from Google currently leads the MMMU-Pro leaderboard with a score of 0.836 across 49 evaluated AI models.
Gemini 3.5 Flash leads with 83.6%, followed by 
GPT-5.5 at 83.2% and GPT-5.4 at 81.2%.
Progress Over Time
Interactive timeline showing model performance evolution on MMMU-Pro
MMMU-Pro Leaderboard
| Context | Cost | License | ||||
|---|---|---|---|---|---|---|
| 1 | Google | — | 1.0M | $1.50 / $9.00 | ||
| 2 | OpenAI | — | 1.1M | $5.00 / $30.00 | ||
| 3 | OpenAI | — | 1.0M | $2.50 / $15.00 | ||
| 3 | Google | — | 1.0M | $0.50 / $3.00 | ||
| 5 | Google | — | — | — | ||
| 6 | Google | — | 1.0M | $2.50 / $15.00 | ||
| 7 |  Meta | — | — | — | ||
| 8 |  Moonshot AI | 1.0T | 262K | $0.95 / $4.00 | ||
| 9 | OpenAI | — | 400K | $1.75 / $14.00 | ||
| 10 |  Alibaba Cloud / Qwen Team | — | 1.0M | $0.50 / $3.00 | ||
| 11 | Moonshot AI | 1.0T | — | — | ||
| 12 | OpenAI | — | — | — | ||
| 13 |  Anthropic | — | 1.0M | $5.00 / $25.00 | ||
| 14 | Google | 31B | 262K | $0.14 / $0.40 | ||
| 14 | Alibaba Cloud / Qwen Team | 122B | 262K | $0.40 / $3.20 | ||
| 16 | Google | — | 1.0M | $0.25 / $1.50 | ||
| 17 | OpenAI | — | 400K | $0.75 / $4.50 | ||
| 18 | OpenAI | — | — | — | ||
| 19 | OpenAI | — | 400K | $5.00 / $30.00 | ||
| 20 | Alibaba Cloud / Qwen Team | 28B | 262K | $0.60 / $3.60 | ||
| 21 | Anthropic | — | 200K | $3.00 / $15.00 | ||
| 22 | Alibaba Cloud / Qwen Team | 35B | — | — | ||
| 23 | Alibaba Cloud / Qwen Team | 35B | 262K | $0.25 / $2.00 | ||
| 24 | Alibaba Cloud / Qwen Team | 27B | 262K | $0.30 / $2.40 | ||
| 25 | Google | 25B | 262K | $0.13 / $0.40 | ||
| 26 | Alibaba Cloud / Qwen Team | 236B | 262K | $0.45 / $3.49 | ||
| 27 | Alibaba Cloud / Qwen Team | 236B | 262K | $0.30 / $1.49 | ||
| 27 | Alibaba Cloud / Qwen Team | 33B | — | — | ||
| 29 | OpenAI | — | 400K | $0.20 / $1.25 | ||
| 30 | Alibaba Cloud / Qwen Team | 33B | — | — | ||
| 31 | Alibaba Cloud / Qwen Team | 31B | — | — | ||
| 32 | Alibaba Cloud / Qwen Team | 9B | 262K | $0.18 / $2.09 | ||
| 32 | Alibaba Cloud / Qwen Team | 31B | — | — | ||
| 34 |  Mistral AI | 119B | 256K | $0.15 / $0.60 | ||
| 35 | OpenAI | — | 128K | $2.50 / $10.00 | ||
| 36 | Meta | 400B | — | — | ||
| 37 | Alibaba Cloud / Qwen Team | 4B | 262K | $0.10 / $1.00 | ||
| 38 | Alibaba Cloud / Qwen Team | 9B | 262K | $0.08 / $0.50 | ||
| 39 | Alibaba Cloud / Qwen Team | 4B | 262K | $0.10 / $0.60 | ||
| 40 | Google | 8B | — | — | ||
| 41 | Alibaba Cloud / Qwen Team | 72B | — | — | ||
| 42 | Alibaba Cloud / Qwen Team | 34B | — | — | ||
| 43 | Alibaba Cloud / Qwen Team | 73B | — | — | ||
| 44 | 90B | — | — | |||
| 45 | Google | 5B | — | — | ||
| 46 |  Microsoft | 6B | — | — | ||
| 47 | Alibaba Cloud / Qwen Team | 8B | — | — | ||
| 48 | Alibaba Cloud / Qwen Team | 7B | — | — | ||
| 49 | 11B | — | — |
FAQ
Common questions about MMMU-Pro.
More evaluations to explore
Related benchmarks in the same category
A challenging dataset of 448 multiple-choice questions written by domain experts in biology, physics, and chemistry. Questions are Google-proof and extremely difficult, with PhD experts reaching 65% accuracy.
A more robust and challenging multi-task language understanding benchmark that extends MMLU by expanding multiple-choice options from 4 to 10, eliminating trivial questions, and focusing on reasoning-intensive tasks. Features over 12,000 curated questions across 14 domains and causes a 16-33% accuracy drop compared to original MMLU.
All 30 problems from the 2025 American Invitational Mathematics Examination (AIME I and AIME II), testing olympiad-level mathematical reasoning with integer answers from 000-999. Used as an AI benchmark to evaluate large language models' ability to solve complex mathematical problems requiring multi-step logical deductions and structured symbolic reasoning.
Massive Multitask Language Understanding benchmark testing knowledge across 57 diverse subjects including STEM, humanities, social sciences, and professional domains
A verified subset of 500 software engineering problems from real GitHub issues, validated by human annotators for evaluating language models' ability to resolve real-world coding issues by generating patches for Python codebases.
Humanity's Last Exam (HLE) is a multi-modal academic benchmark with 2,500 questions across mathematics, humanities, and natural sciences, designed to test LLM capabilities at the frontier of human knowledge with unambiguous, verifiable solutions