Google is rolling out a powerful new AI capability for its most advanced subscribers, and it’s called Deep Think. Starting today, users of Google AI Ultra can access this feature directly in the Gemini app. Far from being a simple upgrade to Gemini 2.5, Deep Think is derived from the same AI variant that achieved top-tier scores at the International Mathematical Olympiad (IMO) — a prestigious competition where elite student mathematicians tackle extremely challenging problems.
While the IMO-winning version of the model required hours to solve such complex challenges, the new Deep Think has been tuned for speed and everyday usability while retaining serious reasoning capabilities. According to Google’s internal testing, the model is still capable of tackling multi-layered, high-stakes problems in a way that is both accurate and efficient.
Deep Think introduces what Google calls parallel thinking. Unlike traditional AI models that work linearly toward a solution, this method allows the system to explore multiple possible solutions at once. It can compare different reasoning paths, revise its approach midway, and synthesize the best answer from several possibilities.
This makes Deep Think especially well-suited for creative and technical challenges such as:
-
Complex coding problems that require detailed analysis of time complexity.
-
Scientific research involving multi-step hypothesis testing.
-
Design projects with several potential solution pathways.
The company also noted that the model now achieves Bronze-level performance on the 2025 IMO benchmark, making it not only fast but also deeply capable in high-level reasoning tasks.
Endorsement from the Top
Google CEO Sundar Pichai took to X (formerly Twitter) to share the news, writing:
“We’re bringing a version of Deep Think that achieved gold-medal status at IMO to Ultra subscribers in the @Geminiapp… Toggle it on when reasoning through complex scientific literature, tackling a coding problem that requires careful consideration of time complexities — or anything else @DemisHassabis considers a fun Friday night.”
His message underlined the versatility of the tool for both professional and personal problem-solving.
How to Access Deep Think
The feature is currently available through the Gemini app on mobile devices for AI Ultra subscribers. To use it:
-
Select the 2.5 Pro model in the app.
-
Enable the Deep Think toggle.
Initially, users will have a limited number of Deep Think prompts per day, but Google plans to expand usage limits and integrate the feature into the Gemini API for selected developers and researchers.
Stronger Safety and Content Controls
Alongside the performance boost, Google says it has improved safety filters and content objectivity. The new model is reportedly better at avoiding harmful or biased outputs, though some testers have noted that it occasionally refuses even harmless queries to err on the side of caution.
The Road Ahead
With Deep Think, Google is positioning Gemini not just as a conversational assistant, but as a serious reasoning partner for high-complexity work. By merging IMO-level logic with user-friendly speed, the company is giving its top-tier users a tool that could reshape how they approach problem-solving, coding, and research.
