MathArena (ETH Zurich): leaderboard

Tests LLMs on recent uncontaminated math olympiad problems (IMO, USAMO, AIME, HMMT, Project Euler, and more). Uses item-response theory. 4 runs per problem with cost tracking.

Source: matharena.ai.

Benchmark results

Results are published separately for each benchmark or evaluation setting.

Interactive version: theaggregate.ai/benchmark?slug=matharena-eth-zurich · How It Works · Data refreshed daily, snapshot 2026-10-20.