AIME 2025

text
+
+
+
+
About

The AIME 2025 benchmark tests AI models using 15 complex mathematical problems from the 2025 American Invitational Mathematics Examination. This rigorous evaluation measures advanced mathematical reasoning, problem-solving, and logical inference across algebra, geometry, and number theory. Each answer is an integer from 000-999, requiring precise step-by-step solutions. The benchmark represents one of the most challenging tests for AI mathematical capabilities.

+
+
+
+
Evaluation Stats
Total Models48
Organizations11
Verified Results0
Self-Reported48
+
+
+
+
Benchmark Details
Max Score1
Language
en
+
+
+
+
Performance Overview
Score distribution and top performers

Score Distribution

48 models
Top Score
100.0%
Average Score
68.8%
High Performers (80%+)
20

Top Organizations

#1Zhipu AI
1 model
93.9%
#2xAI
5 models
93.6%
#3Alibaba Cloud / Qwen Team
7 models
77.9%
#4Anthropic
6 models
77.0%
#5OpenAI
7 models
76.7%
+
+
+
+
Leaderboard
48 models ranked by performance on AIME 2025
LicenseLinks
Jul 9, 2025
Proprietary
100.0%
Oct 15, 2025
Proprietary
96.3%
Aug 7, 2025
Proprietary
94.6%
Sep 30, 2025
MIT
93.9%
Feb 17, 2025
Proprietary
93.3%
Apr 16, 2025
Proprietary
92.7%
Jul 25, 2025
Apache 2.0
92.3%
Aug 28, 2025
Proprietary
92.0%
Jul 9, 2025
Proprietary
91.7%
Aug 7, 2025
Proprietary
91.1%
Showing 1 to 10 of 48 models
...
+
+
+
+
Resources