Google Launches Flagship Gemini 2.0 AI Model Designed for the “Agentic Era”

Google has just announced Gemini 2.0, the company’s most advanced AI model that will power AI overviews in Google Search, new research prototypes, and AI agents. The company says Gemini 2.0 is designed for the “agentic era,” and the flagship AI model will be rolling out to more Google products next year.

After Microsoft launched custom Copilot in public preview last month and OpenAI promised 12 days of announcements in December, Google probably needed to show it’s still firmly engaged in the generative AI race. And today, the search giant detailed several research prototypes leveraging Gemini 2.0 including AI agents for developers and gamers.

“No product has been transformed more by AI than Search,” Google said today. “Our AI Overviews now reach 1 billion people, enabling them to ask entirely new types of questions — quickly becoming one of our most popular Search features ever. As a next step, we’re bringing the advanced reasoning capabilities of Gemini 2.0 to AI Overviews to tackle more complex topics and multi-step questions, including advanced math equations, multimodal queries and coding.”

While Google plans to bring AI Overviews to more languages and countries next year, the company is making the experimental Gemini 2.0 Flash model available for developers and Gemini and Gemini Advanced users today. This new model, which will be generally available in January, can be accessed from the model dropdown menu on the Gemini web app on desktop and mobile. It brings performance improvements and new multimodal capabilities.

“Gemini 2.0 Flash’s native user interface action-capabilities, along with other improvements like multimodal reasoning, long context understanding, complex instruction following and planning, compositional function-calling, native tool use and improved latency, all work in concert to enable a new class of agentic experiences,” Google explained today.

Project Astra, an experimental universal AI assistant that Google announced at its I/O 2024 conference in May is also being upgraded with Gemini 2.0. The latest version can now converse in multiple languages, use Google Search, Lens, and Maps, and remember more conversations. It’s still available via a waitlist for users age 18+ based in the US and the UK.

Google’s other research projects leveraging Gemini 2.0 include Project Mariner, a Chrome extension that works as a research assistant, and an experimental AI-powered code agent named Jules that integrates with GitHub. Google is also exploring how to use Gemini 2.0 to create agents that can help gamers in real time based on what’s happening on the screen.

Lastly, Google is also launching today Deep Research, a new agentic feature available in English in Gemini Advanced on the desktop and mobile web. This feature leverages the existing Gemini 1.5 Pro model, and it lets the chatbot search the web and create in-depth reports on complex topics with source links. Gemini Advanced can try out Deep Research by selecting it from the Gemini Advanced model drop-down menu.