Subscribe

IMO 2025 gold-medal-level result (experimental reasoning model)

An experimental OpenAI reasoning model solves 5 of 6 IMO 2025 problems in natural language with no tools, 35/42 points, a gold-medal score.

Result came from general-purpose reinforcement learning and test-time compute scaling, not a math-specific system or formal verification. Proofs were graded by three former IMO medalists engaged by OpenAI, not IMO coordinators. Google DeepMind reported the same 35/42, officially graded. The model was not released.

Date
Saturday 19 July 2025
Lab
OpenAI
Kind
paper
Access
research preview

Figures

MeasureValueMeasured by
IMO 2025 score35/42 (5 of 6 problems)
proofs graded by three former IMO medalists, unanimous consensus; problem 6 unsolved
mixed
Google DeepMind Gemini Deep Think, same exam35/42
officially graded; for comparison
independent

OpenAI announced the result before the IMO-coordinated release window, and it was graded outside the IMO process. According to Wikipedia, Terence Tao and Kevin Buzzard questioned whether it is comparable to human contestants' results. OpenAI's account comes from a post by researcher Alexander Wei, who said no model with this math capability would ship for months.

Sources

  1. simonwillison.net/2025/Jul/19/openai-gold-medal-math-olympiad/
  2. simonwillison.net/2025/Jul/21/gemini-imo/
  3. github.com/aw31/openai-imo-2025-proofs

This record was checked against its sources on 6 October 2026. How we check

Related