Google is advancing its AI capabilities with the recent release of Gemini 3 Deep Think, designed for tackling intricate scientific, mathematical, and engineering problems. Unlike traditional language models, Deep Think emphasizes complex multi-step reasoning, akin to “System 2” thinking. The update is currently available to Google AI Ultra subscribers and select researchers via the Gemini API. Notable achievements include impressive benchmark scores, such as 48.4% on Humanity’s Last Exam and gold-level performance in the 2025 International Math Olympiad. Deep Think excels in detecting logical errors in technical manuscripts, enhancing its value for academic fields. Practical applications are highlighted through its ability to analyze hand-drawn sketches and produce 3D-printable files. This model embodies a shift toward deliberate AI, stressing logical accuracy and adaptability, thus positioning itself as a vital tool for engineers and scientists. As Google expands access, Deep Think signifies the future of AI, prioritizing intelligence over mere conversation.
Source link
Share
Read more