Google Opens Gemini 4 Carbon to All Developers and AI Ultra Subscribers

Google made Gemini 4 Carbon available to developers, enterprises, scientists and consumers on Tuesday, November 10, ending a short period in which its strongest model could be used only by cybersecurity defenders. Google calls Carbon its strongest model, with its best results in cybersecurity, long agent runs, complex knowledge work, software engineering and scientific reasoning.

The wider release comes two weeks after the first one. On October 28, Google released Carbon to a set of cybersecurity defenders through its Fairwind Program. Carbon follows the recent launch of Gemini 4 Argon. Google says Argon remains the right choice for most software engineering and knowledge work, while Carbon is built for tasks that test the limits of a model, such as complex multi-step coding and open-ended scientific research. Google says Carbon uses more tokens than Argon to explore a problem in depth and run longer checks before it answers.

Where Carbon is available, and what it costs

Carbon is available through the API in Google AI Studio and in Gemini for business for all paying developers and enterprises. The price at launch is the same introductory price as Argon, at 2 dollars per million input tokens and 10 dollars per million output tokens, Google said. 9to5Google reported the introductory price runs until November 30, and rises after that to 4 dollars per million input tokens and 20 dollars per million output tokens.

The model is also rolling out to all Google AI Ultra subscribers in Antigravity, the Gemini app and AI Mode in Search. In the Gemini app model picker, Argon is listed for code and analysis and Carbon for the most complex work, 9to5Google reported. The app has removed Gemini 3.1 Pro with this launch.

Teams of agents, and work on dense files

To show what Carbon can do on larger projects, Google Antigravity is previewing a feature called /swarm. Google says it turns a single prompt into a coordinated team of agents that a developer can track in real time. Developers can give Carbon the roles that need the deepest reasoning, such as architectural design, coordination and auditing, and pair it with faster, lighter models in parallel workspaces, with other agents testing each change.

For legal, finance and enterprise work, Google says Carbon follows strict schemas and formats to produce spreadsheets, presentations and reports from large sets of documents and data. Google says Carbon ranks first on CursorBench 4.0, on Perplexity's WANDR benchmark and on Zapier's AutomationBench, at more than twice lower cost per task than the next best model from another company. Those rankings are Google's own published claims.

Cybersecurity results and safety numbers

Through the Fairwind Program, Google says Carbon leads on SEC-bench Pro for finding vulnerabilities, on ExploitGym for exploiting them and on CWE-bench for patching them. Google said a team at Wiz, working in its Scan for Good program, used Carbon to find zero-day vulnerabilities at two major software as a service providers in under two hours, work Google said can take weeks.

Google also published safety results. In simulated cyberattack tests, Google says Carbon safeguards blocked 99 percent of attack paths from its adaptive red team agent, and cut the per-attempt breach rate to 0.8 percent, down from 3.1 percent on Argon.

Google is also widening science access. Four weeks ago it opened applications to give 10,000 scientists free access to Argon in Antigravity for six months. Carbon access now extends to current and future participants in that program, Google says, and applications remain open.