Definition
Plain language
A hard high school math competition used to test AI math reasoning.
As stated in the literature
The Harvard-MIT Mathematics Tournament, a high-difficulty competition whose problems are used as a reasoning benchmark for language models.
Why it matters: Its difficulty makes it a demanding test of whether a model can do genuine multi-step mathematical reasoning.
For example, an AI might be asked to solve a tricky HMMT olympiad problem that even strong high school competitors find hard.
Heard on the show
“A four-billion-parameter language model — Qwen3-point-five-4B — scores seventy-four percent on HMMT February, which is a real, hard math olympiad.”Episode 071 — When the Model Is Fine and the Plumbing Is Broken: Fixing Agents at the Interface