Hedge funds cut NFLX, keep big bets on MSFT, AMZN, add NVDA
Investing.com -- Google (NASDAQ:GOOGL) DeepMind announced that an advanced version of its Gemini Deep Think AI system has achieved gold-medal standard performance at the 2025 International Mathematical Olympiad (IMO).
The AI system solved five out of six problems perfectly, earning 35 out of 42 possible points, meeting the gold medal threshold. This represents a significant improvement over last year’s silver medal performance when Google’s combined AlphaProof and AlphaGeometry 2 systems scored 28 points.
IMO President Prof. Dr. Gregor Dolinar confirmed the achievement, stating: "Google DeepMind has reached the much-desired milestone, earning 35 out of a possible 42 points — a gold medal score. Their solutions were astonishing in many respects. IMO graders found them to be clear, precise and most of them easy to follow."
Unlike previous AI attempts that required experts to translate problems into specialized languages, this version of Gemini operated entirely in natural language, producing mathematical proofs directly from official problem descriptions within the competition’s 4.5-hour time limit.
The breakthrough utilized an advanced version of Gemini Deep Think with enhanced reasoning capabilities, including "parallel thinking" that allows the model to explore multiple solution paths simultaneously. The system was trained using novel reinforcement learning techniques and given access to high-quality mathematics problem solutions.
Google plans to make this Deep Think model available to trusted testers, including mathematicians, before releasing it to Google AI Ultra subscribers.
Google CEO Sundar Pichai described the progress in mathematical reasoning as "just astounding," while Google DeepMind CEO Demis Hassabis called it "incredible progress."
The announcement comes as OpenAI also reported similar success, with its experimental AI model achieving gold medal-level performance at the same competition with identical scores of 35/42 points.
This article was generated with the support of AI and reviewed by an editor. For more information see our T&C.