Today we are announcing Gemini 2.0 – our most powerful AI model yet, built for the agent age. Gemini 2.0 has new features such as multimodal output with native image generation and audio output, as well as native use of tools such as Google Search and Maps.

We’re releasing an experimental version of Gemini 2.0 Flash, our working model with low latency and improved performance. Developers can start building this model in the Gemini API via Google AI Studio and Vertex AI. And Gemini and Gemini Advanced users around the world can try out a chat-optimized version of Gemini 2.0 by selecting it from the Model drop-down menu on the desktop.

We are also using Gemini 2.0 in new research prototypes, including Project Astra, which explores the future capabilities of a general-purpose AI assistant; Project Mariner, an early prototype that can perform actions in Chrome as an experimental extension; and Jules, an experimental AI-powered code agent. We continue to value security and responsibility in these projects, which is why we take an exploratory and incremental approach to development, including collaboration with trusted testers.

This week we started testing Gemini 2.0 in AI Overviews in Search, and we’ll be expanding Gemini 2.0 to other Google products early next year.