# Google launches Gemini 2.0 with new agentic features and broader developer access
Google says Gemini 2.0 is its most capable AI model yet, and that it is being positioned as the foundation for a new wave of agent-style tools. In a company post dated 2024-12-11, Google said it was releasing the first model in the Gemini 2.0 family and opening parts of the system to developers, trusted testers, and Gemini users.
The headline change is not just raw model quality. Google says Gemini 2.0 is built for an “agentic” era, meaning it can understand more of the world, think several steps ahead, and take action on a user’s behalf with supervision. The company says the model supports native image and audio output, tool use, and improved performance on multimodal tasks.
Google also highlighted a few immediate products. The company said a Gemini 2.0 Flash experimental model would be available to all Gemini users, that a new Deep Research feature would help compile reports on complex topics, and that AI Overviews in Search would begin testing Gemini 2.0 reasoning for harder queries. The release also mentions a Multimodal Live API, native image generation, and text-to-speech features for developers.
The broader strategic message is familiar: Google wants its models to be useful across its core products and not just in a standalone chat interface. The company says Gemini already powers work across its services, and it presents Gemini 2.0 as a step toward a universal assistant that can reason, plan, and act inside different workflows.
The post also tries to ground the release in infrastructure. Google says Gemini 2.0 was trained and run on its Trillium TPUs, and that the company’s custom hardware stack remains central to performance and deployment. That detail matters because AI model launches are increasingly judged not only by benchmark claims, but by how well companies can ship them at scale.
For developers, the practical takeaway is that Google is widening access while keeping the system in an experimental posture. That is a familiar pattern for frontier AI releases: announce capabilities now, limit some functions to early users, then roll them outward as feedback comes in.
The company’s own framing is notably expansive. It says the next phase of Gemini is about making information more useful, not just more available. Whether that promise holds will depend on how the model behaves in real products, but Google is clearly signaling that it sees the “agentic” AI market as the next frontier.
Claim map
The supplied Google blog post supports the claims that Gemini 2.0 launched on Dec. 11, that it adds multimodal output and native tool use, and that Google is rolling out Flash experimental access, Deep Research, and early Search integrations.
