Google is enhancing its AI capabilities with the launch of Deep Think, an advanced reasoning feature for its Gemini app, exclusively available to Google AI Ultra subscribers. Deep Think leverages advanced parallel thinking to tackle complex problems, providing more strategic and creative AI responses. Building on the success of the Gemini 2.5 model, which excelled at the 2025 International Mathematical Olympiad, Deep Think achieves “Bronze-level IMO performance,” optimized for quicker, practical use.
This feature allows Gemini to extend its “thinking time,” exploring multiple hypotheses and refining solutions, which significantly boosts its ability to handle intricate tasks like coding, scientific reasoning, and algorithmic development. Google envisions Deep Think transforming workflows for developers and researchers, particularly in formulating mathematical conjectures and delving into complex literature.
Currently, Deep Think has surpassed performance of other models on notable benchmarks and integrates seamlessly with existing Gemini features. Google prioritizes safety, implementing rigorous evaluations to mitigate risks associated with advanced AI reasoning.
Source link
