Google has made Gemini 4 Carbon available to developers today through the API's paid tier in Android Studio and Google AI Studio. Google AI Ultra subscribers can also try the model through Google Antigravity and AI Studio, giving subscribers a separate route to access it.
Google says Gemini 4 Carbon is designed for agents working through complicated tasks over extended periods. Its intended uses include scientific research, building full-stack websites and software engineering, with the company also emphasizing its cybersecurity abilities.
Read more: Google's Gemini Robotics ER 2 Opens in Public Preview on Cloud
The developer opening follows an introduction on October 28, when Google scheduled initial access for selected cyber defenders and planned to extend it gradually through its Fairwind Program to additional trusted partners. At that introduction, Google set introductory pricing at $2 per million input tokens and $10 per million output tokens, alongside a 95% discount for cached input tokens.
Google describes Gemini 4 Carbon as its most capable model and claims improvements in alignment and its ability to carry tasks through to completion. The company positions those capabilities as support for developers creating and shipping secure products.
Developers can consult the Gemini API Developer Guide for setup help and details about the model. Google also directs users to the Google Antigravity blog for instructions on working with Gemini 4 Carbon in that development platform, which is built around agents.
