Google Releases Gemini 2.0, Its Most Advanced AI Model: Here’s What’s New

Google has announced the launch of Gemini 2.0, its most advanced and capable generative AI model yet. The company said that the launch of Gemini 2.0 will usher in a new “agentic era.” According to CEO Sundar Pichai, with new advances in multimodality—like native image and audio output—Gemini 2.0 will help build new AI agents with a vision of a universal assistant.

Starting Dec. 11, Gemini 2.0 will be available for developers and testers, and the Gemini 2.0 Flash experimental model will be available to all Gemini users.

Here’s what’s new in the Gemini 2.0 release:

The first model in the Gemini 2.0 family is Gemini 2.0 Flash. In addition to supporting multimodal inputs like images, video, and audio, 2.0 Flash now supports multimodal output like natively generated images mixed with text and steerable text-to-speech multilingual audio. It can also natively call tools like Google Search, code execution, as well as third-party user-defined functions.

According to Google, 2.0 Flash is twice as fast as Gemini 1.5 Pro and outperforms it on key benchmarks.

“It’s our workhorse model with low latency and enhanced performance at the cutting edge of our technology, at scale,” said Demis Hassabis, CEO of Google DeepMind.

2.0 Flash is available now as an experimental model to developers via the Gemini API in Google AI Studio and Vertex AI. General availability will be in January.

Google’s AI Overviews enable users to ask questions and get an AI overview based on the subject of search. Now, Google is bringing advanced reasoning capabilities of Gemini 2.0 to AI Overviews. This will help in handling more complex topics and multi-step questions, including advanced math equations, multimodal queries and coding.

The company has started limited testing this week and will be rolling it out more broadly next year. AI Overviews will also be available to more countries and languages over the next year, Google said.

The new feature uses advanced reasoning and long context capabilities to act as a research assistant. For instance, if a user wants to write a report on a complex topic, Deep Research can browse the topic on the web and compile a report on behalf of the user.

Deep Research is available in Gemini Advanced.

The universal virtual assistant announced during Google I/O has been improved in the release of Gemini 2.0. Project Astra can now converse in multiple languages and in mixed languages, with a better understanding of accents and uncommon words.

The assistant can now use Google Search, Lens and Maps to help users. It also now has up to 10 minutes of in-session memory and can remember more user conversations.

According to Hassabis, the research prototype “explores the future of human-agent interaction, starting with your browser.” It’s able to understand and reason across information in your browser screen, including pixels and web elements like text, code, images and forms, and then uses that information via an experimental Chrome extension to complete tasks for you.

The AI-powered code agent for developers can tackle an issue, develop a plan and execute it under developer supervision. Jules is integrated directly into GitHub workflow.