A Message from Sundar Pichai, CEO of Google and Alphabet

Information is the foundation of human progress. For over 26 years, we’ve been dedicated to our mission of organizing the world’s information and making it accessible and useful. Today, we’re pushing the boundaries of AI to organize information across diverse inputs and make it accessible through various outputs—helping it become even more valuable for you.

That vision guided us when we introduced Gemini 1.0 last December. As the first natively multimodal AI model, Gemini 1.0 and its successor, Gemini 1.5, made significant advancements in understanding and processing information across text, video, images, audio, and code, with an impressive ability to handle long contexts.

Millions of developers are now building with Gemini, helping us reimagine existing products—including our seven with over 2 billion users—and creating entirely new ones. NotebookLM, for instance, showcases the potential of multimodality and long context, becoming a favorite tool for many.

Over the past year, we’ve invested heavily in creating more agentic AI models—systems that understand the world around you, think multiple steps ahead, and take action on your behalf, all with your oversight.

Today, we’re thrilled to introduce Gemini 2.0, our most advanced model yet, ushering in the era of agentic AI. With breakthroughs in multimodality, including native image and audio outputs, and enhanced tool integration, Gemini 2.0 paves the way for the development of AI agents aligned with our vision of a universal assistant.

Starting today, Gemini 2.0 is available to developers and trusted testers. Its capabilities are also rolling out across our products, beginning with Gemini and Search. Additionally, the Gemini 2.0 Flash experimental model is now accessible to all Gemini users, alongside a new feature called Deep Research, which leverages advanced reasoning and long context to act as a powerful research assistant, available in Gemini Advanced.

Search has been transformed more by AI than any other product. Our AI Overviews now serve 1 billion users, enabling them to ask new types of questions. With Gemini 2.0’s advanced reasoning, we’re enhancing AI Overviews to handle complex topics, multi-step queries, advanced math, multimodal inputs, and coding challenges. Limited testing is underway, with a broader rollout planned for early next year, along with expanded language and country support.

Gemini 2.0 is built on the foundation of our decade-long investments in AI innovation, powered by Trillium, our sixth-generation TPUs, which drove 100% of its training and inference. Starting today, Trillium is also available for customers to use in their own projects.

If Gemini 1.0 was about organizing and understanding information, Gemini 2.0 is about making that information truly useful. I’m excited to see how this next chapter unfolds.