Back to Benchmarks
🧠

MMLU-Pro

MMLU-Professional

12K graduate-level questions across 14 domains, 10 options each

🧠 Knowledge & Reasoning
Best: 90%
18 Models Evaluated
Champion

GPT-oss 120B

OpenAI

90%

Best Score

Model Rankings

All models ranked by score on MMLU-Pro. Scores are relative to the best score of 90%.

#ModelProviderScorevs Best
1GPT-oss 120B
Elo 1355
OpenAI
90%
Best
2Claude Opus 4.7
Elo 1500
Anthropic
88%
-2.0%
3GPT-5.5
Elo 1478
OpenAI
88%
-2.0%
4Qwen 3.5
Elo 1450
Alibaba
87.8%
-2.2%
5DeepSeek V4 Pro (Max)DeepSeek
87.5%
-2.5%
6Kimi K2.5
Elo 1438
Moonshot
87.1%
-2.9%
7Kimi K2 ThinkingMoonshot
87.1%
-2.9%
8Step-3.5-Flash
Elo 1389
Stepfun
85.8%
-4.2%
9Gemini 3.1 Pro
Elo 1492
Google
85%
-5.0%
10Gemini 3 Pro
Elo 1486
Google
85%
-5.0%
11DeepSeek V3.2
Elo 1423
DeepSeek
85%
-5.0%
12MiMo-V2-Flash
Elo 1393
Xiaomi
84.9%
-5.1%
13DeepSeek R1
Elo 1398
DeepSeek
84%
-6.0%
14Claude Opus 4.6
Elo 1503
Anthropic
82%
-8.0%
15Llama 4 Maverick
Elo 1328
Meta
80.5%
-9.5%
16Claude Sonnet 4.6
Elo 1460
Anthropic
79.1%
-10.9%
17MiniMax M2.5
Elo 1404
MiniMax
76.5%
-13.5%
18GLM-5
Elo 1454
Zhipu AI
70.4%
-19.6%

Score Distribution

GPT-oss 120B
90%
Claude Opus 4.7
88%
GPT-5.5
88%
Qwen 3.5
87.8%
DeepSeek V4 Pro (Max)
87.5%
Kimi K2.5
87.1%
Kimi K2 Thinking
87.1%
Step-3.5-Flash
85.8%
Gemini 3.1 Pro
85%
Gemini 3 Pro
85%

Run This Benchmark

Learn how to evaluate your own models on MMLU-Pro.

MMLU-Pro is a standardized benchmark for evaluating AI model capabilities.

To run this benchmark on your own models, visit the official benchmark page or leaderboard for detailed instructions, code, and evaluation scripts.

Data compiled from official model cards, public leaderboards, and independent evaluations.