AI Research Atlas

AlphaProof and AlphaGeometry 2 (IMO 2024 silver)

Google DeepMind · 25 July 2024

AlphaProof plus AlphaGeometry 2 solved 4 of 6 IMO 2024 problems for 28/42 points, silver-medal level.

AlphaProof does RL over formal Lean proofs; AlphaGeometry 2 is Gemini-based with a 100x faster symbolic engine, solving 83% of historical IMO geometry problems vs 53%. Problems were formalised by hand and some took up to three days, far beyond the 4.5-hour limit.

Date
Thursday, 25 July 2024
Lab
Google DeepMind
Kind
paper
Access
research preview

Figures

MeasureValueMeasured by
IMO 2024 score28/42
gold threshold 29; 4 of 6 problems
third-party

Graded by Fields Medallist Timothy Gowers and Joseph Myers per DeepMind. Superseded by Gemini Deep Think gold in natural language (2025-07-21).

Sources

  1. deepmind.google/blog/ai-solves-imo-problems-at-silver-medal-level/

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

Related