Google Launches ‘Deep Think’ in Gemini App: IMO-Level AI Reasoning Now in Your Pocket

In a significant upgrade for its power users, Google has rolled out “Deep Think”—a powerful new AI feature integrated into the Gemini app. Exclusively available to Google AI Ultra subscribers, Deep Think is designed to elevate reasoning and problem-solving capabilities, bringing Olympiad-level mathematical intelligence directly to your device.

From Mathematical Olympiads to Your Smartphone

The Deep Think model isn’t just a routine update to Gemini 2.5. Instead, it's built on a variant that achieved Bronze-level performance in the 2025 International Mathematical Olympiad (IMO) benchmark—a prestigious global competition for elite student mathematicians.

While the IMO-version of this model required hours to process intricate problems, the Gemini app edition is optimized for responsiveness, ensuring it's suitable for everyday use while retaining advanced reasoning capabilities.

“It’s a significant improvement over what was first announced at I/O,” said Google in its announcement, referring to the Gemini 2.5 demo unveiled at its annual developer event.

What Makes Deep Think Unique: Parallel Thinking Explained

The core of Deep Think’s power lies in its “parallel thinking” capability. Rather than following a linear approach to solving a problem, the AI evaluates multiple possibilities simultaneously. This makes it adept at:

• Exploring different solutions paths

• Comparing trade-offs between methods

• Revising strategies mid-task

According to Google, this approach makes Deep Think particularly effective for “creative or layered tasks — things like complex coding challenges, scientific research, and step-by-step design projects.”

Performance Benchmarks: IMO-Level Intelligence, Optimized for Real Use

With its release in Gemini, Deep Think achieves Bronze-level performance on the 2025 IMO benchmark, according to internal Google evaluations. While it doesn’t yet match the exact capabilities of its gold-medal variant, it now strikes a balance between high-level reasoning and practical responsiveness.

Sundar Pichai’s Announcement

Google CEO Sundar Pichai shared the news on X (formerly Twitter), stating:

“We’re bringing a version of Deep Think that achieved gold-medal status at IMO to Ultra subscribers in the @Geminiapp (+ the official version is now in the hands of mathematicians). 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.”

How to Access Deep Think in the Gemini App

For users eager to try it out, here’s how to get started:

Activation Guide:

1. Open the Gemini app on your smartphone.

2. Select the 2.5 Pro model.

3. Toggle on the “Deep Think” option.

Note: At launch, access is limited to a set number of Deep Think prompts per day. However, Google plans to broaden availability gradually, including access through the Gemini API for researchers and developers.

Focus on Safety and Responsibility

Alongside its powerful reasoning capabilities, Deep Think also features enhanced content safety filters. Google reports that it has made strides in improving objectivity and content moderation compared to earlier AI versions.

Still, the system may err on the side of caution, occasionally declining harmless prompts in an effort to maintain safe usage boundaries.

What This Means for the Future of AI Interaction

The release of Deep Think marks an exciting leap in AI usability. Users can now interact with a system that combines advanced academic-level reasoning with real-world usability. Whether you're dissecting scientific papers, optimizing algorithms, or designing complex systems, Deep Think can act as a powerful co-thinker, not just a tool.

With further plans to expand its access and capabilities, this rollout reaffirms Google’s commitment to intelligent, responsible, and creative AI.

Conclusion 

Google’s rollout of the Deep Think feature in the Gemini app is a groundbreaking step in making advanced AI reasoning more accessible. With roots in a variant that earned Bronze-level performance at the 2025 International Mathematical Olympiad, Deep Think delivers high-level problem-solving abilities in a mobile-optimized format.

Its use of parallel thinking allows it to evaluate multiple possibilities at once, making it ideal for complex tasks such as scientific research, in-depth coding, and structured design. By limiting usage initially and focusing on content safety, Google aims to ensure a responsible rollout.

As users begin to explore its capabilities, Deep Think is expected to enhance how people approach intellectually demanding problems—acting as a co-pilot in reasoning, rather than a tool. With API access coming for developers and researchers, Deep Think could soon become a cornerstone for advanced AI interaction, bridging the gap between Olympiad-grade intelligence and everyday problem-solving.