AIster · The agentic age · Breakthroughs
AI models reach gold-medal standard at the Maths Olympiad
General language models wrote olympiad-level proofs in plain English, within exam time
Key facts
- Date
- 19–21 Jul 2025
- Era
- The agentic age
- Interest
- Breakthroughs
- Threads
- Model wars
- Score
- 35 of 42 points, five of six problems solved
- Who
- OpenAI (experimental reasoning model) and Google DeepMind (Gemini Deep Think)
- Grading
- Google: official IMO coordinators. OpenAI: three former IMO medallists
- Announced
- OpenAI on Saturday 19 July; Google on Monday 21 July 2025
At the 2025 International Mathematical Olympiad, an experimental OpenAI model and an advanced version of Google DeepMind's Gemini Deep Think each solved five of the six problems, scoring 35 out of 42, enough for a gold medal. Both worked in natural language within the two 4.5-hour exam sessions. Google's result was graded by IMO coordinators; OpenAI's was graded by three former IMO medallists it hired.
Sources
- Google DeepMind, Advanced version of Gemini with Deep Think officially achieves gold-medal standard at the International Mathematical Olympiad, 2025-07-21
- TechCrunch, OpenAI and Google outdo the mathletes, but not each other, 2025-07-21
- The Decoder, OpenAI claims a breakthrough in LLM reasoning on complex math problems, 2025-07-20