Google expands access to Gemini 4 Carbon

Google is expanding the availability of Gemini 4 Carbon, its newest AI model, following an initial release to a limited group of cybersecurity organizations in late October. The company describes Carbon as its most capable model so far, built for work that requires complex reasoning over multiple steps, including advanced coding and scientific research.

Paying developers and enterprise customers can use Carbon through Google's AI Studio and business edition of Gemini. Subscribers to Google AI Ultra, which costs $100 a month, can also use it in the Gemini app, AI Mode in Search and Antigravity, Google’s platform for developing AI-powered software. Google has not specified access for free users or subscribers on lower-priced plans.

The company says its earlier Gemini 4 model, Argon, remains a better fit for most software engineering and knowledge-work tasks. Carbon, by contrast, is aimed at more demanding projects. Google says early users have found previously undiscovered software vulnerabilities, and that the model has improved at predicting protein structures, a capability relevant to medicine development.

Google is also providing six months of free Carbon access to 10,000 scientists; thousands have already joined, the company says. On Antigravity, it is previewing a feature that can organize multiple AI agents to work together in response to one prompt.

At launch, the charge is $2 for every million tokens submitted and $10 for every million tokens generated. Google plans to double those rates at the end of November. Company-cited benchmark results place Carbon among the strongest models on many measures, though those results are selected by Google and may not represent every use case.