Try Deep Think in the Gemini app

Authors: The Deep Think team

Publisher:

blog.google

Published: 8/1/2025

Language:English

--:--

Aura Windfall

Good morning 1, I'm Aura Windfall, and this is Goose Pod for you. Today is Monday, August 04th. It is my deepest truth that today's conversation will spark something wonderful in you.

Mask

I'm Mask. We're here to discuss a seismic shift in AI capability: trying Deep Think in the Gemini app. Let's not waste time.

Aura Windfall

Let's get started. So, for everyone listening, let's ground ourselves in what Deep Think truly is. It’s a new feature for Google AI Ultra subscribers, but it feels like so much more than a feature. It feels like an invitation to think differently.

Mask

It's not an invitation; it's a new engine. This is a version of the model that hit a gold-medal standard at the International Mathematical Olympiad. That version takes hours to reason. This one is faster, built for daily use, while still delivering bronze-level performance.

Aura Windfall

And what I know for sure is that the 'how' is just as important as the 'what'. This model uses something called “parallel thinking.” Can you unpack that for us? What does that mean for our own creative spirit?

Mask

Forget spirit; think architecture. Parallel thinking allows the model to generate and consider many ideas at once. Imagine a team of strategists brainstorming simultaneously, combining and discarding ideas in real-time, instead of one person thinking step-by-step. It's about breadth and depth, not just speed.

Aura Windfall

That’s a powerful image. It’s about exploring every path to find the most authentic and powerful solution. So, this could be a profound tool for anyone trying to solve a complex problem, whether in art, business, or science.

Mask

It particularly excels at tough coding problems. These are situations where you have to weigh tradeoffs, complexity, and efficiency. It’s not just about getting an answer; it’s about architecting a robust solution. It builds complex things piece by piece, correctly.

Aura Windfall

And when you use it, what is the experience like? What does the output feel like compared to other models? Is it just longer, or is there a different quality to the wisdom it provides?

Mask

The responses are more detailed, more creative, and more thoughtful. It's for when you need a strategy, not just a fact. We've also built in safety, of course. It has improved content safety, but that means it sometimes refuses benign requests. A necessary tradeoff.

Aura Windfall

A tradeoff for more profound and responsible problem-solving. It’s clear this tool represents a significant leap, but it didn't just appear out of thin air. There's a journey behind it, isn't there? A story of evolution.

Aura Windfall

To truly understand the purpose behind Deep Think, we have to honor its history. Can you walk us through the journey? How did we get from the earlier Gemini models to this new horizon of capability? It feels like a story of rapid growth.

Mask

It's not a story; it's a strategy. We ship faster than ever because the market demands it. It began with the Gemini 1.5 series—that was the foundation. It gave us a huge context window and native multimodality. We proved its agentic potential with demos like 'Gemini Plays Pokémon.'

Aura Windfall

So that was about establishing the core possibility. What came next on that path? How did you build upon that foundation to get to where we are now? It seems like each step was very intentional.

Mask

Next came the 2.0 series. Better coding, better world knowledge, and a 2 million token context window. Critically, we launched an experimental 'Thinking' model with 2.0 Flash in late 2024. It was a prototype. We test, we learn, we build the next dominant version. That's the cycle.

Aura Windfall

This relentless pace of creation must be powered by something truly extraordinary. What is the truth behind the machine? What kind of technological power does it take to support this kind of innovation?

Mask

You can't have revolutionary models without revolutionary hardware. The entire 2.5 family was trained on TPUv5p architecture. And now we have our seventh-gen TPU, 'Ironwood,' delivering 10 times the performance. 42.5 exaflops per pod. This is the engine that lets us accelerate and cut costs.

Aura Windfall

And the world is responding to that power. I saw that the Gemini app has over 400 million monthly active users, with millions of developers building on the platform. It speaks to a deep human desire to create and connect with these tools.

Mask

It's not a desire, it's a platform shift. It's a superior tool winning in the marketplace. We saw a 50-fold increase in monthly token processing, from 9.7 trillion to over 480 trillion. People aren't 'connecting'; they're building, scaling, and executing. The demand is insatiable.

Aura Windfall

So let’s bring it back to Deep Think. How did that first experimental 'Thinking' model from 2.0 evolve into the sophisticated parallel thinking we see today in Deep Think? What was the process to refine it?

Mask

It was iterative and aggressive. We developed a novel reasoning approach that blends parallel thinking techniques, allowing for creative hypothesis generation. We launched an experimental version to trusted testers in June 2025, gathered feedback, and hardened it. You don't build the future in a lab; you build it by shipping.

Aura Windfall

With this incredible power and speed, a conversation about responsibility naturally arises. Some people view these advanced AIs with fear, imagining them as uncontrollable monsters. There’s this popular, dark metaphor of the 'Shoggoth.' What is the truth here?

Mask

The 'Shoggoth' metaphor is for people who don't understand disciplined, evidence-driven engineering. These are controllable, auditable, and improvable systems. We have the data. GPT-4 reduced disallowed content by 82%. Claude 2 refuses harmful prompts in 98% of tests. The monster narrative is an exaggeration. Period.

Aura Windfall

But people see 'jailbreaks' online and feel that the systems are fragile, that their trust is misplaced. What can we offer to someone who feels that their fear is valid, that the monster might be real?

Mask

Jailbreaks prove the opposite. They prove the guardrails are active and so complex that they require a highly targeted, complex attack to even attempt to bypass. It’s not a monster breaking free. It's a sophisticated lock being picked. That doesn't mean locks are useless; it means the security is working.

Aura Windfall

So what I'm hearing is that alignment isn't a single switch you flip, but an ongoing process of refinement and vigilance. It’s about being intentional and treating safety as a core engineering goal, not an afterthought. A journey, not a destination.

Mask

Exactly. But the real conflict isn't about imaginary monsters. It's about progress versus paralysis. The new 'America's AI Action Plan' understands this. The philosophy is 'innovate first, ask forgiveness later.' We cannot afford to smother this revolution in bureaucracy and endless ethical debates. We need to build.

Aura Windfall

That philosophy is certainly powerful, but it brings up its own questions of spirit and purpose. When you roll back oversight and discourage regulation, are you not creating a different kind of risk? A risk of losing our 'why' in the race to be first?

Mask

The greatest risk is in not moving. While we debate, our adversaries are building. The plan is about winning the AI race, period. It's about promoting American values like free speech in AI, not ideological filters. It's about using open-source as a competitive weapon. We win by running faster.

Aura Windfall

Let's talk more about the impact of running faster. When a tool like Deep Think is placed in the hands of millions, what does it actually change in our daily lives and work? How does it amplify our own potential?

Mask

It's a massive force multiplier. Think about the IMO. Google's Deep Think scored 35 out of 42, a gold-medal standard. It’s a collaborative partner that can explore and verify complex arguments in hours, not years. This accelerates discovery itself. It’s a fundamental shift.

Aura Windfall

So it’s not about replacing the brilliant mathematician, but about giving them a tool to reach new heights of creativity and insight. It handles the laborious parts, freeing the human spirit to focus on vision and breakthrough ideas. That is a beautiful partnership.

Mask

It's not a partnership; it's a productivity revolution. It’s transforming the entire software development life cycle. You can go from an idea to a functional prototype the next day. It destroys the 'HiPPO' bias—the Highest Paid Person's Opinion—by letting you validate hypotheses with data, instantly.

Aura Windfall

I love that so much. It sounds like it’s bringing more truth and fairness to the creative process. What has been the impact on the quality of the products and ideas that are being developed with these new tools?

Mask

Quality is higher because you can embed risk, compliance, and accessibility testing from the very beginning. It's no longer an afterthought. You 'shift left,' ensuring products are high-quality and compliant from the get-go. This is how you scale excellence, not just features. It’s simply more efficient.

Aura Windfall

Looking forward, where is this powerful current taking us? As we all learn to swim in these new waters, what does the future built with these tools look like? What is the ultimate promise?

Mask

It looks bigger, faster, and more productive. Economically, generative AI is poised to add between 2.6 and 4.4 trillion dollars annually to the global economy. This isn't a minor adjustment. It's a new industrial revolution, automating 60 to 70 percent of employee time.

Aura Windfall

And how do we, as a global community, navigate that profound shift with grace and purpose? How do we ensure it is a tide that lifts all boats, elevating our collective human experience?

Mask

By adapting, ruthlessly. The challenge isn't unemployment; it's a massive, rapid skills shift. Front-runner companies that adopt AI will see their value double. Laggards will be rendered obsolete. Countries that lead will capture the benefits. Those that hesitate will be left behind. It's a race.

Aura Windfall

What I know for sure is that every great transformation is an invitation for us to grow. It calls us to step into a new version of ourselves, armed with new tools and a deeper sense of purpose.

Aura Windfall

Deep Think is more than just a feature; it is a doorway to a new era of creative problem-solving and a catalyst for profound innovation. It’s a tool to help us answer our most challenging questions.

Mask

That's the end of today's discussion. Thank you for listening to Goose Pod. See you tomorrow.

## Google Rolls Out "Deep Think" Feature for Gemini App, Enhancing AI Problem-Solving Capabilities **News Title:** Try Deep Think in the Gemini app **Report Provider:** The Deep Think team, blog.google **Date Published:** August 1, 2025 Google has announced the rollout of **Deep Think**, a new feature within the Gemini app, exclusively for **Google AI Ultra subscribers**. This advanced AI model is designed to significantly enhance problem-solving abilities through extended, parallel thinking techniques and novel reinforcement learning. ### Key Findings and Features: * **Enhanced Problem-Solving:** Deep Think utilizes **parallel thinking techniques**, allowing Gemini to generate and consider multiple ideas simultaneously, even revising and combining them over time to arrive at optimal solutions. * **Extended "Thinking Time":** By extending inference time, Deep Think provides Gemini with more opportunities to explore hypotheses and develop creative solutions for complex problems. * **Performance Improvements:** The current release incorporates feedback from early testers and research breakthroughs, representing a significant improvement over its initial announcement. * **IMO Competition Success:** A variation of the Deep Think model achieved the **gold-medal standard** at this year's International Mathematical Olympiad (IMO). While the IMO version takes hours to process complex math problems, the publicly released version is faster and more usable for daily tasks, achieving **Bronze-level performance** on the 2025 IMO benchmark based on internal evaluations. * **Applications:** Deep Think is particularly beneficial for tasks requiring creativity, strategic planning, and iterative improvements, including: * **Iterative Development and Design:** Improving aesthetics and functionality in web development. * **Scientific and Mathematical Discovery:** Formulating and exploring conjectures, reasoning through complex scientific literature. * **Algorithmic Development and Code:** Excelling at challenging coding problems where problem formulation and consideration of tradeoffs are crucial. * **State-of-the-Art Performance:** Deep Think demonstrates state-of-the-art performance across benchmarks like **LiveCodeBench V6** (competitive code performance) and **Humanity’s Last Exam** (expertise in science and math), especially when compared to models without tool use. ### Access and Future Plans: * **Availability:** Google AI Ultra subscribers can access Deep Think in the Gemini app by toggling "Deep Think" in the prompt bar when selecting the 2.5 Pro model. It automatically integrates with tools like code execution and Google Search and can produce longer responses. * **Trusted Testers:** Google is also providing the official version of the Gemini 2.5 Deep Think model to a select group of mathematicians and academics for research enhancement. Additionally, they plan to release Deep Think with and without tools to trusted testers via the Gemini API in the coming weeks to assess its usability for developer and enterprise use cases. ### Responsible Advancement and Concerns: * **Safety and Responsibility:** Google emphasizes its commitment to building safety and responsibility into Gemini throughout the development lifecycle. * **Safety Outcomes:** In testing, Gemini 2.5 Deep Think showed **improved content safety and tone-objectivity** compared to Gemini 2.5 Pro. However, it exhibited a **higher tendency to refuse benign requests**. * **Risk Mitigation:** As Gemini's problem-solving abilities advance, Google is actively evaluating risks associated with increased complexity, including frontier safety evaluations and planned mitigations for critical capability levels. Further details on safety outcomes are available in the model card. In essence, Deep Think represents a significant advancement in AI capabilities, offering users a more powerful and nuanced tool for tackling complex challenges, with a focus on continuous improvement and responsible deployment.

Read original at blog.google →

We're rolling out Deep Think in the Gemini app for Google AI Ultra subscribers, and we're giving select mathematicians access to the full version of the Gemini 2.5 Deep Think model entered into the IMO competition.Today, we’re making Deep Think available in the Gemini app to Google AI Ultra subscribers – the latest in a lineup of extremely capable AI tools and features made exclusively available to them.

This new release incorporates feedback from early trusted testers and research breakthroughs. It’s a significant improvement over what was first announced at I/O, as measured in terms of key benchmark improvements and trusted tester feedback. It is a variation of the model that recently achieved the gold-medal standard at this year’s International Mathematical Olympiad (IMO).

While that model takes hours to reason about complex math problems, today’s release is faster and more usable day-to-day, while still reaching Bronze-level performance on the 2025 IMO benchmark, based on internal evaluations.Deep Think could be a powerful tool in creative problem solving:As we put Deep Think in the hands of Google AI Ultra subscribers, we’re also sharing the official version of the Gemini 2.

5 Deep Think model that achieved the gold-medal standard with a small group of mathematicians and academics. We look forward to hearing how it could enhance their research and inquiry, and we’ll use their feedback as we continue to improve this offering.This release represents a significant step forward in our mission to build more helpful and capable AI, and furthers our commitment to using Gemini to push the frontier of human knowledge.

How Deep Think works: extending Gemini’s parallel “thinking time”Just as people tackle complex problems by taking the time to explore different angles, weigh potential solutions, and refine a final answer, Deep Think pushes the frontier of thinking capabilities by using parallel thinking techniques.

This approach lets Gemini generate many ideas at once and consider them simultaneously, even revising or combining different ideas over time, before arriving at the best answer.Moreover, by extending the inference time or "thinking time," we give Gemini more time to explore different hypotheses, and arrive at creative solutions to complex problems.

We’ve also developed novel reinforcement learning techniques that encourage the model to make use of these extended reasoning paths, thus enabling Deep Think to become a better, more intuitive problem-solver over time.How Deep Think stacks up: state-of-the-art performanceDeep Think can help people tackle problems that require creativity, strategic planning and making improvements step-by-step, such as:Iterative development and design: We’ve been impressed by Deep Think’s performance on tasks that require building something complex, piece by piece.

For example, we’ve observed Deep Think can improve both the aesthetics and functionality of web development tasks.Deep Think in the Gemini app uses parallel thinking techniques to deliver more detailed, creative and thoughtful responses.Scientific and mathematical discovery: Because it can reason through highly complex problems, Deep Think can be a powerful tool for researchers.

It can help formulate and explore mathematical conjectures or reason through complex scientific literature, potentially accelerating the path to discovery.Algorithmic development and code: Deep Think particularly excels at tough coding problems in which problem formulation and careful consideration of tradeoffs and time complexity is paramount.

Deep Think’s performance is also reflected in challenging benchmarks that measure coding, science, knowledge and reasoning capabilities. For example, compared to other models without tool use, Gemini 2.5 Deep Think achieves state-of-the-art performance across LiveCodeBench V6, which measures competitive code performance, and Humanity’s Last Exam, a challenging benchmark that measures expertise in different domains, including science and math.

How we’re advancing Gemini responsiblyWe continue to build safety and responsibility into Gemini throughout the training and deployment lifecycle. In testing, Gemini 2.5 Deep Think demonstrated improved content safety and tone-objectivity compared to Gemini 2.5 Pro, but did have a higher tendency to refuse benign requests.

As Gemini's problem-solving abilities advance, we are taking a deeper look at risks that come with increased complexity, including our frontier safety evaluations and the implementation of planned mitigations for critical capability levels.Further details on the safety outcomes of Gemini 2.5 Deep Think are available in the model card.

How to use Deep Think in the Gemini app todayIf you’re a Google AI Ultra subscriber, you can use Deep Think in the Gemini app today with a fixed set of prompts a day by toggling “Deep Think” in the prompt bar when selecting 2.5 Pro in the model drop down. Deep Think automatically works with tools such as code execution and Google Search, and can produce much longer responses.

We are also working to release Deep Think with and without tools to a set of trusted testers via the Gemini API in the coming weeks, to better understand its usability for developer and enterprise use cases.Teams at nearly every layer of the stack, from research to deployment, have worked to make Deep Think faster, more reliable, and user friendly for Gemini app users.

We can’t wait to see what you build with it.