MMLU
Massive Multitask Language Understanding benchmark testing knowledge across 57 diverse subjects including STEM, humanities, social sciences, and professional domains
GPT-5 from OpenAI currently leads the MMLU leaderboard with a score of 0.925 across 99 evaluated AI models.
GPT-5 leads with 92.5%, followed by o1 at 91.8% and o1-preview at 90.8%.
Progress Over Time
Interactive timeline showing model performance evolution on MMLU
MMLU Leaderboard
| Context | Cost | License | ||||
|---|---|---|---|---|---|---|
| 1 | OpenAI | — | — | — | ||
| 2 | OpenAI | — | — | — | ||
| 3 | OpenAI | — | — | — | ||
| 3 | OpenAI | — | — | — | ||
| 5 |  Sarvam AI | 105B | — | — | ||
| 5 |  Alibaba Cloud / Qwen Team | 236B | — | — | ||
| 7 |  Anthropic | — | — | — | ||
| 7 | Anthropic | — | — | — | ||
| 9 |  Moonshot AI | 1.0T | — | — | ||
| 9 | OpenAI | — | 1.0M | $2.00 / $8.00 | ||
| 11 | OpenAI | 117B | 131K | $0.09 / $0.45 | ||
| 12 |  Meituan | 560B | 128K | $0.30 / $1.20 | ||
| 13 | Moonshot AI | 1.0T | — | — | ||
| 13 | Moonshot AI | 1.0T | — | — | ||
| 15 | Alibaba Cloud / Qwen Team | 236B | 262K | $0.30 / $1.49 | ||
| 16 | Alibaba Cloud / Qwen Team | 33B | — | — | ||
| 16 | OpenAI | — | 128K | $2.50 / $10.00 | ||
| 18 |  DeepSeek | 671B | — | — | ||
| 19 | Alibaba Cloud / Qwen Team | 235B | — | — | ||
| 20 | Moonshot AI | 1.0T | — | — | ||
| 21 | Alibaba Cloud / Qwen Team | 31B | — | — | ||
| 22 |  xAI | — | — | — | ||
| 22 | OpenAI | — | 1.0M | $0.40 / $1.60 | ||
| 24 | Moonshot AI | — | — | — | ||
| 25 | 405B | — | — | |||
| 26 | OpenAI | — | — | — | ||
| 27 | Anthropic | — | — | — | ||
| 28 | OpenAI | — | 128K | $10.00 / $30.00 | ||
| 29 | Alibaba Cloud / Qwen Team | 33B | — | — | ||
| 29 | OpenAI | — | — | — | ||
| 31 | xAI | — | — | — | ||
| 32 | 90B | — | — | |||
| 32 | 70B | — | — | |||
| 34 |  Google | — | — | — | ||
| 34 |  Amazon | — | — | — | ||
| 36 | OpenAI | — | 128K | $2.50 / $10.00 | ||
| 37 | Meituan | 69B | 256K | $0.10 / $0.40 | ||
| 38 |  Meta | 400B | — | — | ||
| 39 | OpenAI | 21B | — | — | ||
| 40 | OpenAI | — | — | — | ||
| 40 | Alibaba Cloud / Qwen Team | 9B | 262K | $0.18 / $2.09 | ||
| 42 | Sarvam AI | 30B | — | — | ||
| 43 | Alibaba Cloud / Qwen Team | 31B | — | — | ||
| 44 |  Microsoft | 15B | — | — | ||
| 45 |  Mistral AI | 123B | — | — | ||
| 46 | 70B | — | — | |||
| 47 | Alibaba Cloud / Qwen Team | 33B | — | — | ||
| 48 | Alibaba Cloud / Qwen Team | 72B | — | — | ||
| 49 | OpenAI | — | — | — | ||
| 50 | Alibaba Cloud / Qwen Team | 4B | 262K | $0.10 / $1.00 |
FAQ
Common questions about MMLU.
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.
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
LiveCodeBench is a holistic and contamination-free evaluation benchmark for large language models for code. It continuously collects new problems from programming contests (LeetCode, AtCoder, CodeForces) and evaluates four different scenarios: code generation, self-repair, code execution, and test output prediction. Problems are annotated with release dates to enable evaluation on unseen problems released after a model's training cutoff.