Supports· weakORCH-E024
Gemini Deep Think: officially graded IMO gold, in natural language
- what the source states
- DeepMind reports that an advanced version of Gemini with Deep Think solved five of six problems at the 2025 International Mathematical Olympiad for 35 of 42 points, achieving gold-medal standard. The announcement states the solutions were graded and certified by IMO coordinators using the same criteria applied to student contestants, and that the system worked end-to-end in natural language within the 4.5-hour competition limit — without the Lean formalization used in 2024. The page attributes a confirming statement to IMO President Gregor Dolinar.
- what we infer
- Included for two specific facts the Nature paper predates: official IMO grading under contest conditions, and natural-language rather than formalized operation. Graded weak as a source because it is a corporate blog post with no accompanying paper and an unreleased model, so it rests on DeepMind's account — though the load-bearing element, external grading by IMO coordinators, is the part least susceptible to self-report bias. Same scoping as the rest of this theme: motivation rung only.
- limitations
- Corporate announcement; no peer-reviewed paper and no released model.
- Solutions in natural language were not machine-verified, unlike the 2024 Lean work.
- Competition mathematics, not open research problems.
- Bears on the motivation rung only.