HMMT 2025
text
+
+
+
+
About
HMMT 2025 is a mathematical reasoning benchmark based on problems from the 2025 Harvard-MIT Mathematics Tournament, one of the world's most prestigious high school mathematics competitions. This benchmark tests AI models' ability to solve olympiad-level mathematical problems requiring advanced techniques in algebra, geometry, combinatorics, and number theory, measuring elite mathematical problem-solving capabilities.
+
+
+
+
Evaluation Stats
Total Models10
Organizations3
Verified Results0
Self-Reported10
+
+
+
+
Benchmark Details
Max Score1
Language
en
+
+
+
+
Performance Overview
Score distribution and top performers
Score Distribution
10 models
Top Score
93.3%
Average Score
59.5%
High Performers (80%+)
3Top Organizations
#1DeepSeek
3 models
65.5%
#2OpenAI
5 models
64.1%
#3Moonshot AI
2 models
38.8%
+
+
+
+
Leaderboard
10 models ranked by performance on HMMT 2025
| License | Links | ||||
|---|---|---|---|---|---|
| Aug 7, 2025 | Proprietary | 93.3% | |||
| Aug 7, 2025 | Proprietary | 87.8% | |||
| Sep 29, 2025 | MIT | 83.6% | |||
| May 28, 2025 | MIT | 79.4% | |||
| Aug 7, 2025 | Proprietary | 75.6% | |||
| Sep 5, 2025 | MIT | 38.8% | |||
| Jul 11, 2025 | MIT | 38.8% | |||
| Apr 14, 2025 | Proprietary | 35.0% | |||
| Jan 10, 2025 | MIT | 33.5% | |||
| Apr 14, 2025 | Proprietary | 28.9% |
