DeepMind’s AlphaGeometry solves 25 of 30 olympiad geometry problems
See how this story connectsGoogle DeepMind published AlphaGeometry in Nature on 17 January 2024. The system, which pairs a language model with a symbolic deduction engine and was trained on 100 million synthetic examples, solved 25 of 30 past olympiad geometry problems within time limits, against 25.9 for an average human gold medallist and 10 for the previous best method. DeepMind released the code and model.
Why it matters
Researchers got open code for an olympiad-level geometry prover trained without human proofs, and DeepMind built AlphaGeometry 2 from it for the July 2024 International Mathematical Olympiad.
- When
- 17 Jan 2024
- ID
WTA-N-2024-0019- Where
- United Kingdom, United States
- Who
- Google DeepMind
- Topics
- Mathematics, Open-weight models
Summary by wheresthe.ai, from Google DeepMind and the reporting credited above. Checked against the source on 9 Oct 2026.
Source copy: download the page as saved on 18 Oct 2026 (PDF). Also held by the Internet Archive.
Cite this story
WTA-N-2024-0019: “DeepMind’s AlphaGeometry solves 25 of 30 olympiad geometry problems.” wheresthe.ai, 17 Jan 2024. https://wheresthe.ai/d/deepmind-alphageometry-2024-01/ (source: Google DeepMind).
Spotted an error? Email [email protected].