MMMLU

Multilingual
text
+
+
+
+
About

MMMLU (Multilingual Massive Multitask Language Understanding) extends the MMLU benchmark to multiple languages, evaluating language models' knowledge and reasoning capabilities across diverse linguistic contexts. This multilingual adaptation tests models' ability to understand and apply academic knowledge in various languages, assessing both linguistic competency and cross-cultural knowledge transfer.

+
+
+
+
Evaluation Stats
Total Models14
Organizations4
Verified Results0
Self-Reported14
+
+
+
+
Benchmark Details
Max Score1
Language
en
+
+
+
+
Performance Overview
Score distribution and top performers

Score Distribution

14 models
Top Score
89.5%
Average Score
81.4%
High Performers (80%+)
10

Top Organizations

#1Anthropic
5 models
88.0%
#2Alibaba Cloud / Qwen Team
1 model
86.7%
#3OpenAI
6 models
81.2%
#4Microsoft
2 models
62.7%
+
+
+
+
Leaderboard
14 models ranked by performance on MMMLU
LicenseLinks
Aug 5, 2025
Proprietary
89.5%
Sep 29, 2025
Proprietary
89.1%
May 22, 2025
Proprietary
88.8%
Dec 17, 2024
Proprietary
87.7%
Apr 14, 2025
Proprietary
87.3%
Apr 29, 2025
Apache 2.0
86.7%
May 22, 2025
Proprietary
86.5%
Feb 24, 2025
Proprietary
86.1%
Feb 27, 2025
Proprietary
85.1%
Aug 6, 2024
Proprietary
81.4%
Showing 1 to 10 of 14 models
+
+
+
+
Resources