Google has announced Gemini 4 Hydrogen for developer tools and consumer products, positioning it for coding and agent tasks. Developers can choose among four thinking levels, with high set as the default. The rollout spans model access and Google applications, while an Antigravity orchestrator can switch between Hydrogen and Carbon within the same working session.
Google announced Gemini 4 Hydrogen on November 12 for developer tools and consumer products. The company positions the model for everyday coding, tool use and agent tasks involving multiple steps. Access includes the Gemini developer interface, Google AI Studio and applications such as Gemini and Search. The announcement covers a model rollout across those services rather than a change confined to Search.[1], [2]
Developers can choose low, medium, high or extra-high thinking levels, with high identified by Google as the default. The settings govern how much computation the model devotes to a task before answering. Applications can therefore configure that processing effort for their work. Google’s descriptions of coding and agent capabilities are its product assessments, rather than an assurance of identical outcomes on every task.[1]
Google also describes an orchestrator in Antigravity, its development environment, which can switch between Carbon and Hydrogen within the same session. This permits different models to handle parts of a continuing workflow. Paid Antigravity subscribers are among the announced access routes. Hydrogen is distributed through Google services; the announcement does not release its model weights or provide an open-source software licence.[1]
