Artificial intelligence · 2025

LLMs win IMO gold

General-purpose language models solve five of six olympiad problems in natural language.

At the July 2025 International Mathematical Olympiad, an experimental OpenAI reasoning model and Google DeepMind's Gemini Deep Think each scored 35 of 42 points, gold medal level, writing proofs in plain English within the 4.5-hour limits. A year earlier the best system needed formal Lean and days of compute.

See it in the living treeWikipedia