Maths
AIME 2025
The American Invitational Mathematics Examination — 15 problems, integer answers, no partial credit. A standard test of multi-step maths reasoning.
AIME 2025 is scored as a percentage of questions answered correctly. 56 of the 159 models in this index report a score for it. The highest published figure here is 96%, from Claude Opus 4.5.
Reported scores
Top 20 on AIME 2025
Ordered by the figure each maker published. Models that have not reported this benchmark are not listed — an absent score is not a low score.
| # | Model | Published score |
|---|---|---|
| 1 | Claude Opus 4.5Anthropic | |
| 2 | Gemini 3 ProGoogle DeepMind | |
| 3 | GPT-5OpenAI | |
| 4 | Kimi K2 ThinkingMoonshot AI | |
| 5 | Grok 4xAI | |
| 6 | GLM-4.6Z.ai (Zhipu) | |
| 7 | o4-miniOpenAI | |
| 8 | gpt-oss-120bOpenAI | |
| 9 | Qwen3-MaxAlibaba Qwen | |
| 10 | Grok 4 FastxAI | |
| 11 | Seed-OSS-36BByteDance Seed | |
| 12 | GPT-5 miniOpenAI | |
| 13 | GLM-4.5Z.ai (Zhipu) | |
| 14 | Grok 3 MinixAI | |
| 15 | gpt-oss-20bOpenAI | |
| 16 | GLM-4.5-AirZ.ai (Zhipu) | |
| 17 | DeepSeek-V3.2DeepSeek | |
| 18 | o3OpenAI | |
| 19 | DeepSeek-V3.1DeepSeek | |
| 20 | o3-miniOpenAI |
Where these numbers come from
Every score on this page is a published figure, taken from the model's own card, system card, technical report or release post, or from a public leaderboard. CorX Labs did not run these evaluations. Most are self-reported by the lab that built the model, which means they were produced under that lab's own choice of prompt, scaffold and number of attempts — so treat them as a starting point for a shortlist, not as a settled ranking.
A score someone other than the model's maker measured is marked Independent and names its measurer. Those are the stronger numbers on this page — an outside harness has no reason to flatter anyone — and there are not many of them.
Where a figure has not been published, the cell reads Not reported rather than an estimate. Nothing here is inferred, interpolated or guessed. Each model records the month its row was last checked. Full method and caveats.
How to read this score
AIME problems have a single integer answer and no partial credit, so the score is unambiguous. It is the standard proxy for multi-step symbolic reasoning, and the benchmark where reasoning models separated themselves most sharply from the rest.
Where it is weak
Fifteen problems a year makes for a very small sample, and labs often report the average of many attempts rather than a single pass, which is not the same thing. Competition maths also has little in common with the arithmetic in ordinary work.