📐

AIME 2025

American Invitational Math Exam

30 problems from 2025 AIME, integer answers 000-999

📐 Math

Best: 100%

21 Models Evaluated

Champion

Claude Opus 4.6

Anthropic

100%

Best Score

All models ranked by score on AIME 2025. Scores are relative to the best score of 100%.

#	Model	Provider	Score	vs Best
1	Claude Opus 4.6 Elo 1503	Anthropic	100%	Best
2	Gemini 3.1 Pro Elo 1492	Google	100%	+0.0%
3	Step-3.5-Flash Elo 1389	Stepfun	99.8%	-0.2%
4	Kimi K2 Thinking	Moonshot	99.1%	-0.9%
5	Claude Opus 4.7 Elo 1500	Anthropic	99%	-1.0%
6	GPT-5.5 Elo 1478	OpenAI	99%	-1.0%
7	GPT-5.4 Elo 1463	OpenAI	99%	-1.0%
8	GPT-5.4 Pro Elo 1480	OpenAI	99%	-1.0%
9	GPT-oss 120B Elo 1355	OpenAI	97.9%	-2.1%
10	Kimi K2.5 Elo 1438	Moonshot	96.1%	-3.9%
11	Gemini 3 Pro Elo 1486	Google	96%	-4.0%
12	DeepSeek V4 Pro (Max)	DeepSeek	95.2%	-4.8%
13	MiMo-V2-Flash Elo 1393	Xiaomi	94.1%	-5.9%
14	Grok 3 Elo 1412	xAI	93.3%	-6.7%
15	DeepSeek V3.2 Elo 1423	DeepSeek	89.3%	-10.7%
16	Mistral Large Elo 1416	Mistral	88%	-12.0%
17	DeepSeek R1 Elo 1398	DeepSeek	87.5%	-12.5%
18	MiniMax M2.5 Elo 1404	MiniMax	86.3%	-13.7%
19	GLM-5 Elo 1454	Zhipu AI	84%	-16.0%
20	Nemotron Ultra 253B Elo 1348	NVIDIA	72.5%	-27.5%
21	Claude Sonnet 4.6 Elo 1460	Anthropic	52.8%	-47.2%

Claude Opus 4.6

100%

Gemini 3.1 Pro

100%

Step-3.5-Flash

99.8%

Kimi K2 Thinking

99.1%

Claude Opus 4.7

99%

GPT-5.5

99%

GPT-5.4

99%

GPT-5.4 Pro

99%

GPT-oss 120B

97.9%

Kimi K2.5

96.1%

Learn how to evaluate your own models on AIME 2025.

AIME 2025 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.