Gemini 4 Carbon is now available to paying developers and AI Ultra subscribers

Google made Gemini 4 Carbon available on Nov. 10, two weeks after it limited the model to cyber defenders in an earlier story. Paying developers and enterprises get it through the API, and AI Ultra subscribers are getting it in the Gemini app.

What changed

Google says Carbon is its strongest model, with frontier performance in cybersecurity and long-horizon agentic work such as software engineering and scientific reasoning. Google wrote that it is making Carbon available to "developers, enterprises, scientists, and consumers."

Google says Argon remains "the ideal choice for the majority of software engineering and knowledge work." Carbon is for tasks that test a model's limits, and Google says it uses more tokens than Argon to reason deeply and run verification loops. Our Argon story covers that model.

Who is affected

Developers and enterprises on a paid plan can use Carbon in Google AI Studio and Gemini for business. Google's developer post also names Android Studio.

Google says Carbon is rolling out to all AI Ultra subscribers in Google Antigravity, the Gemini app and AI Mode in Search. 9to5Google reports that Argon stays available alongside it for AI Pro and up.

Google's post does not list Carbon for free or AI Pro users. It does not say they are excluded either.

Google also extended its program giving 10,000 scientists free Argon in Antigravity for six months to Carbon. Scientists apply, with rolling onboarding and eligibility rules. Cyber defenders can apply to the Fairwind Program for "Carbon for cyber."

Price, benchmarks and safety

The API price matches Argon's introductory rate: $2 per million input tokens and $10 per million output tokens. A footnote says that period ends Nov. 30, 2026, after which $4 and $20 apply.

Google says Carbon ranks first on CursorBench 4.0, Perplexity's WANDR and Zapier's AutomationBench, at more than 2X lower cost per task than the next best external model. It says Carbon ranks second on a new Multimodal Knowledge Agent benchmark, at more than 6X lower average cost per task than the top model. These are Google's results.

Google says Carbon blocked 99% of attack trajectories from an adaptive red-teaming agent in simulated cyberattack tests, and its per-attempt breach rate was 0.8%, against 3.1% for Argon. Google says outside tester Trajectory Labs found no new universal jailbreaks across 132 hours and over 13,000 requests. Google also says prompt injection robustness still needs work.

Google also describes a preview of /swarm in Antigravity. It turns one prompt into a team of agents you can track in real time, with Carbon in roles such as architecture and auditing.

What is not known

We do not know whether free or AI Pro users will get Carbon, or what usage limits AI Ultra has. Google gave no date for broader tiers. We found no price beyond the $4 and $20 it stated, and no independent benchmark.

What this means for you

If you pay for AI Ultra, look for Carbon in the Gemini app as it rolls out. If you build with the API, the introductory price ends Nov. 30, so budget for the higher rate. Everyone else should stay on Argon for now. More in AI apps.

Google also says AlphaProof Code used Carbon with other models on Signal protocol code, which we covered in our story.

What we checked

We read Google's Nov. 10 post and developer post, plus a report on the Gemini app rollout, on Nov. 10. We did not test the model or verify any benchmark.