AIME
A hard American high school mathematics competition whose problems are now a standard benchmark for AI reasoning.
AIME stands for American Invitational Mathematics Examination, a competition for high school students who have already done well on an earlier qualifying round. The problems are genuinely difficult, each answer is a whole number between 0 and 999, and there is no partial credit. That combination makes it convenient to grade automatically, which is why AI labs adopted it.
Scores on AIME are now quoted in almost every model launch, and they have climbed fast: several 2026 models report above 99 percent. Treat those numbers with a little care. Because the problems are public and heavily discussed online, there is a permanent worry that models have effectively seen them during training, and labs also tune specifically for famous tests. A high AIME score tells you a model is good at contest mathematics. It does not tell you much about whether it will remember a fact correctly.