Gemini 2.0 is built for AI's "new agentic era," Google chief executive Sundar Pichai said.

The tech giant introduced Gemini 2.0, which “will enable us to build new AI agents that bring us closer to our vision of a universal assistant,” Google chief executive Sundar Pichai said Wednesday, adding that the company is working to roll out the new model across its products. AI agents are software that can complete complex tasks autonomously for a user.

Gemini 2.0 has new multimodal capabilities, including native image and audio output, Pichai said. Google launched Gemini 1.0 last December, which the company said was the first “natively multimodal” model, meaning it could process and respond to text, video, image, audio, and code inquiries.

Developers and testers will be the first to get 2.0, while all Gemini users will have access to the Gemini 2.0 Flash experimental model. The Flash model builds off of Gemini 1.5 Flash, which Google launched in July as its fastest, most cost-efficient model.

Google will add Gemini 2.0's reasoning capabilities to its AI Overviews feature, which Pichai said now reaches one billion people and is “quickly becoming one of our most popular Search features ever.” With Gemini 2.0, AI Overviews will be able to solve advanced multi-step queries, such as mathematical equations and multimodal questions.