Google launches Deep Think reasoning mode for Gemini 2.5 Ultra

Google has launched Deep Think, a new reasoning mode for its Gemini 2.5 Ultra AI model that allows the system to engage in multi-step thinking before responding to complex queries. Available now to Google AI Ultra subscribers ($249/month), the feature represents Google’s latest attempt to compete with advanced reasoning models by giving its AI more “thinking time” to solve problems requiring strategy, iteration, and complex logic.

What you should know: Deep Think transforms how Gemini processes difficult requests by implementing parallel thinking capabilities that mirror human problem-solving approaches.

• Instead of rushing to deliver immediate answers, the AI generates multiple ideas, evaluates them simultaneously, and refines its approach before responding.
• The model has achieved bronze-level performance on the 2025 International Mathematical Olympiad benchmark and outperformed competitors on challenging tests like LiveCodeBench V6 and Humanity’s Last Exam.
• Google collaborated with academic researchers, including mathematician Michel van Garrel, to test Deep Think’s potential for theoretical problems and scientific discovery.

How it works: Ultra subscribers can toggle Deep Think on within the Gemini app when using Gemini 2.5 Pro for more thoughtful, detailed responses.

• The mode supports longer responses and integrates with tools like Google Search and code execution.
• Usage is currently limited to a fixed number of prompts per day.
• While the original research model took hours to solve Olympiad-level math problems, the consumer version is optimized for faster, practical daily use.

Why this matters: Deep Think positions Google to compete more effectively in the advanced AI reasoning space, particularly for applications requiring complex problem-solving like code debugging, website building, and advanced mathematics.

• The feature addresses growing demand for AI systems that can handle multi-step reasoning tasks beyond simple question-answering.
• Google’s emphasis on safety includes improved tone and content moderation, though the system may occasionally refuse benign requests as the company continues frontier safety evaluations.

The big picture: While Deep Think currently remains exclusive to expensive Ultra subscribers, Google’s historical pattern suggests broader accessibility could follow if the feature proves successful and stable.

• Google previously expanded access to powerful features like Gemini Pro and image generation to free-tier users after initial testing phases.
• Broadening Deep Think access would help Google showcase its capabilities and maintain competitiveness against other tech giants in the AI reasoning race.