AIME (benchmark)
AIME, the American Invitational Mathematics Examination, is a 15-question US maths competition with whole-number answers from 0 to 999; AI labs use its past papers, such as AIME 2025, as a maths reasoning benchmark.
Also called: AIME 2025, AIME 2024, American Invitational Mathematics Examination
By Divya Sharma · Checked Nov 15, 2026 · How we review
The name has two meanings. First, AIME is a real exam run by the Mathematical Association of America for students who score well on the AMC 10 or AMC 12. It has 15 questions and three hours in total to answer them, and sits in the path towards the USA Mathematical Olympiad.
AIME as an AI benchmark
Second, in model announcements "AIME 2025" means a model's score on that year's questions. AIME suits AI testing because the problems need several steps of reasoning yet every answer is a single integer from 0 to 999, so grading is automatic and exact. Evaluations such as the Artificial Analysis Intelligence Index have used AIME 2025 with a mathematical equality checker.
Reading the scores
AIME results became a common way to show off reasoning models, which spend extra test-time compute working through a problem before answering. Because a single exam has only 15 questions, one problem moves the score a lot, so small differences between models mean little. Past papers are published online, so older years may have leaked into training data, the risk known as benchmark contamination. That is why reports usually name the year, and the newest year is the most trustworthy.
Limits
AIME tests competition maths with short answers. It says little about proofs, applied maths in code, or explaining a solution clearly. Use it as one signal among benchmarks.
Related terms
- AI benchmark
- An AI benchmark is a fixed, public set of tasks with an agreed scoring method, used to compare models on the same test, for example coding issues solved, exam questions answered or retrieval quality across many datasets.
- Reasoning model
- A reasoning model is a language model trained to work through a problem in a stream of intermediate thinking tokens before giving its final answer, which tends to help on maths, coding and multi-step tasks.
- Benchmark contamination
- Benchmark contamination is when a model has seen a benchmark's test questions or answers during training, so its score on that benchmark overstates how well it handles new problems of the same kind.
- Test-time compute
- Test-time compute is the amount of computation a model spends while answering a request, rather than during training. Letting a model think longer or try several answers can improve results on hard problems.
- Chain-of-thought
- Chain-of-thought (CoT) is a prompting method in which a language model writes out intermediate reasoning steps before its final answer, which tends to improve results on maths, logic and other multi-step problems.