A message from Google and Alphabet CEO Sundar Pichai:
Almost two years ago, we ushered in the Gemini era, one of our largest scientific and product efforts we have ever undertaken as a company. Since then, it has been incredible to see how many people love it. AI Overviews now has 2 billion users monthly. The Gemini app has over 650 million users per month, more than 70% of our Cloud customers use our AI, 13 million developers have built with our generative models, and that is just a fraction of the impact we are seeing.
And we are able to bring advanced capabilities to the world faster than ever, thanks to our differentiated full-stack approach to AI innovation – from our leading infrastructure to our research, our models and world-class tools, to products that reach billions of people around the globe.
Each generation of Gemini has built on the previous one, enabling you to do more. The breakthroughs of Gemini 1 in native multimodality and a long context window expanded the type of information that could be processed and the amount of it. Gemini 2 laid the foundation for 'agentic' (autonomous) capabilities and pushed the boundaries of reasoning and thinking, aiding in more complex tasks and ideas. This led to Gemini 2.5 Pro leading the LMArena leaderboard for more than six months.
And now we introduce Gemini 3, the most intelligent Gemini model that combines all the capabilities of Gemini so you can bring any idea to life.
It is state-of-the-art in reasoning, built to understand depth and nuance, whether it’s perceiving subtle cues in a creative idea or dissecting the overlapping layers of a difficult problem. Gemini 3 is also much better at grasping the context and intent behind your request, so you get what you need with less prompting. It’s amazing to think that AI has evolved in just two years from simply reading text and images to 'reading the room' (understanding social context).
And since November 18, Gemini 3 has been rolled out widely within the Google ecosystem. This includes Gemini 3 in AI Mode in Search with more complex reasoning and new dynamic experiences. This is the first time we are making Gemini available in Search from day one. Gemini 3 is available in the Gemini app, for developers in AI Studio and Vertex AI, and in the new agent development platform: Google Antigravity – more on that later in this article.
Like the generations before it, Gemini 3 again pushes the boundaries of what is possible. In this new chapter, we continue to push the limits of intelligence, agents, and personalization to make AI truly useful for everyone.
Gemini 3: the most intelligent model that helps you bring any idea to life
With the introduction of Gemini 3 on November 18, Google has taken another major step towards AGI.
According to Google, it is the best model in the world for multimodal understanding and the most powerful model for 'agentic' tasks and 'vibe coding' to date. It delivers richer visualizations and deeper interactivity – all built on a foundation of state-of-the-art reasoning.
The Gemini 3 era begins with the preview release of Gemini 3 Pro, which is available in various Google products, allowing you to use it in your daily life to learn, build, and plan everything. Additionally, a separate mode has been launched: Gemini 3 Deep Think, an enhanced reasoning mode that further boosts the performance of Gemini 3. Gemini 3 Deep Think is a mode that will first be made available to safety testers before being rolled out to Google AI Ultra subscribers.
See the following video by clicking this link:
State-of-the-art reasoning with unprecedented depth and nuance
Gemini 3 Pro can bring any idea to life with its advanced reasoning and multimodal capabilities. It performs significantly better than 2.5 Pro on every major AI benchmark.
It tops the LMArena Leaderboard with a groundbreaking score of 1501 points. It demonstrates reasoning at a PhD level with top scores on Humanity's Last Exam (37.5% without using tools) and GPQA Diamond (91.9%). It also sets a new standard for frontier models in mathematics, with a new state-of-the-art of 23.4% on MathArena Apex.
In addition to text, Gemini 3 Pro redefines multimodal reasoning with breakthrough scores of 81% on MMMU-Pro and 87.6% on Video-MMMU. It also achieves a state-of-the-art 72.1% on SimpleQA Verified, showing significant progress in factual accuracy. This means that Gemini 3 Pro is highly capable of solving complex problems across a wide range of subjects such as science and mathematics, with a high degree of reliability.
Gemini 3 Pro also brings a new level of depth and nuance to every interaction. The answers are smart, concise, and direct, trading clichés and flattery for real insight – it tells you what you need to hear, not just what you want to hear. It acts as a true thinking partner, giving you new ways to understand information and express yourself, from translating compact scientific concepts by generating code for high-fidelity visualizations to creative brainstorming sessions.
See the following video by clicking this link:
Gemini 3 can code a visualization of plasma flow in a tokamak and write a poem that captures the physics of fusion.
Gemini 3 Deep Think
The mode Gemini 3 Deep Think pushes the boundaries of intelligence even further, delivering a step-change in Gemini 3's reasoning and multimodal understanding capabilities to help you solve even more complex problems.
In tests, Gemini 3 Deep Think outperforms the already impressive performance of Gemini 3 Pro on Humanity's Last Exam (41.0% without the use of tools) and GPQA Diamond (93.8%). It also achieves an unprecedented 45.1% on ARC-AGI-2 (with code execution), demonstrating its ability to tackle new challenges.
Gemini 3 Deep Think mode excels in some of the most challenging AI benchmarks. More details about the evaluation method can be found in Google's documentation.
Gemini 3 helps you learn, build, and plan everything
Learn everything
Gemini has been built from the ground up to seamlessly integrate information on any topic across multiple modalities, including text, images, video, audio, and code. Gemini 3 pushes the boundary of multimodal reasoning to help you learn in ways that make sense to you, by combining its state-of-the-art reasoning, vision, and spatial awareness, leading multilingual performance, and a context window of 1 million tokens.
For example, if you want to learn to cook according to your family tradition, Gemini 3 can decipher handwritten recipes in different languages and translate them into a shareable family cookbook. Or if you want to learn more about a new topic, you can provide it with academic papers, long video lectures, or tutorials, and it can generate code for interactive flashcards, visualizations, or other formats that help you master the material. It can even analyze videos of your pickleball match, identify areas where you can improve, and generate a training plan for enhancing your overall fitness.
Gemini 3 can help you preserve your family's cooking traditions. Try in Gemini Canvas.
With Gemini 3, you can analyze complex information, such as research papers, and generate code for an interactive guide.
Receive expert-level sports analysis of your pickleball match or other sports to improve your game.
To help you better understand information on the web, AI Mode in Search now uses Gemini 3 to enable new generative UI experiences, such as immersive visual layouts and interactive tools and simulations, all fully generated on the spot based on your search query.
Learn a complex topic such as how RNA polymerase works with generative UI in AI mode in Search (currently only in the US).
Build everything
Building on the success of 2.5 Pro, Gemini 3 delivers on the promise of bringing any idea to life for developers. It excels in zero-shot generation and processes complex prompts and instructions to render richer, more interactive web UIs.
Google positions Gemini 3 as the best 'vibe coding' and 'agentic coding' model to date – it aims to make Google's products more autonomous and increase developer productivity. It tops the WebDev Arena leaderboard with an impressive ELO score of 1487. It also scores 54.2% on Terminal-Bench 2.0, which tests a model's ability to use tools to operate a computer via the terminal, and it performs significantly better than 2.5 Pro on SWE-bench Verified (76.2%), a benchmark that measures coding agents.
You can now build with Gemini 3 in Google AI Studio, Vertex AI, Gemini CLI, and the new agent development platform, Google Antigravity. It is also available on third-party platforms such as Cursor, GitHub, JetBrains, Manus, Replit, and more.
Program a retro 3D spaceship game with rich graphics and enhanced interactivity. Try in AI studio.
Bring your imagination to life by building, deconstructing, and remixing detailed 3D voxel art using code. Try in AI studio.
Build a playable sci-fi world with shaders using Gemini 3. Try in AI studio.
With Gemini 3, you can build richer, more interactive web interfaces and apps through 'vibe-coding'.
Introducing a new 'agent-first' development experience
As model intelligence accelerates with Gemini 3, the opportunity arises to reshape the entire development experience. Since the release of Gemini 3 on November 18, Google Antigravity has been available, the new agent development platform that enables developers to work at a higher, task-oriented level.
By leveraging Gemini 3's advanced reasoning, tool usage, and agentic coding capabilities, Google Antigravity transforms AI assistance from a tool in a developer's toolkit into an active partner. While the core of Google Antigravity is a familiar AI IDE experience, the agents have been elevated to a dedicated surface and have direct access to the editor, terminal, and browser. Now agents can autonomously plan and execute complex, end-to-end software tasks on your behalf while validating their own code.
Alongside Gemini 3 Pro, Google Antigravity is also closely integrated with the latest Gemini 2.5 Computer Use model for browser control and Google's highest-rated image editing model Nano Banana (Gemini 2.5 Image).
See the following video by clicking this link:
Google Antigravity uses Gemini 3 to drive an end-to-end agentic workflow for a flight tracker app. The agent autonomously plans, codes the application, and validates its execution via computer use in the browser.
Plan everything
Since the introduction of the agent era with Gemini 2, Google has made significant progress, not only by improving Gemini's coding agent skills but also by enhancing the ability to reliably plan ahead over a longer horizon. Gemini 3 demonstrates this by leading the leaderboard on Vending-Bench 2, which tests planning over a longer horizon by managing a simulated vending machine business. Gemini 3 Pro maintains consistent tool usage and decision-making for an entire simulated year of operation, ensuring higher returns without deviating from the task.
Chart: Vending-Bench 2: Average over 5 runs per model. Gemini 3 Pro demonstrates better long-term planning to generate significantly higher returns compared to other frontier models.
This means that Gemini 3 can better help you get things done in daily life. By combining deeper reasoning with improved, consistent tool usage, Gemini 3 can take action on your behalf by navigating more complex, multi-step workflows from start to finish – such as triaging your Gmail inbox or planning a travel itinerary – all under your control and guidance.
Google AI Ultra subscribers in the US can now utilize these Gemini agent capabilities in the Gemini app. According to Google, the work on Gemini's agent capabilities has already yielded many insights, and the company is eager to learn about user experiences and feedback as it gradually expands the functionality to more Google products.
Gemini Agent can help you organize your Gmail inbox.
Building Gemini 3 responsibly
Gemini 3 is Google's safest model to date and has undergone the most extensive set of safety evaluations of any Google AI model to date. The model shows reduced sycophancy (talking to please), increased resistance to prompt injections, and improved protection against abuse via cyberattacks.
In addition to internal testing for the critical domains in Google's Frontier Safety Framework, Google has also collaborated on evaluations with globally leading subject matter experts, granted early access to entities such as the UK AISI, and obtained independent reviews from industry experts such as Apollo, Vaultis, Dreadnode, and more. For more information, see the Gemini 3 model card.
The next era of Gemini
This is just the beginning of the Gemini 3 era. Since November 18, Gemini 3 is available to try:
- For everyone in the Gemini app and for Google AI Pro and Ultra subscribers in AI Mode in Search.
- For developers in the Gemini API in AI Studio, Google's new agent development platform Google Antigravity, and Gemini CLI.
- For enterprises in Vertex AI and Gemini Enterprise.
Additionally, Google AI Pro is free for students for 1 year as mentioned in this article. The student offer runs until December 9 and can be ordered here (there you will also find a deeper dive into the benefits for students).
For the Gemini 3 Deep Think mode, Google is taking extra time for safety evaluations and input from safety testers before making it available in the coming weeks for Google AI Ultra subscribers.
Google plans to add more models to the Gemini 3 series soon, allowing users to do more with AI. Google has also indicated that it looks forward to feedback from users and is curious about what they learn, build, and plan with Gemini.
