Technology

Gemini 4 Carbon Rolls Out: Price, Swarm Agents and Who Gets Access

By Editorial Staff ·
Graphic on Google's Gemini 4 Carbon AI model rollout

Google has started rolling out Gemini 4 Carbon, the most powerful model in its new Gemini 4 family, to a much wider group of users. From this week, developers on the paid tier of the Gemini API and subscribers to the Google AI Ultra plan can use it, two weeks after the company first gave access only to a small group of cybersecurity defenders. Google is also previewing a feature called /swarm that turns a single request into a team of AI agents working together.

The wider release was announced on Tuesday, November 10, 2026. It puts Google in a sharper race with OpenAI and Anthropic to offer the most capable models for coding, research and business work.

What Gemini 4 Carbon Is Built For

Carbon sits at the top of the Gemini 4 lineup. Google unveiled it on October 28, about two weeks after it introduced Gemini 4 Argon, according to Reuters. In the Gemini app’s model picker, Argon is labelled “Code and analysis” while Carbon is labelled “Most complex work”.

Google says Carbon is aimed at long, demanding tasks such as complex software projects, scientific research, legal and financial document work, enterprise analysis, and building games and web apps. For most everyday coding and office tasks, the company still recommends Argon, saying it “remains the ideal choice for the majority” of such work.

A strong focus on cybersecurity

When Carbon was first unveiled, Google highlighted its security skills. Reuters reported that Google described the model as state of the art at finding software vulnerabilities, testing them and writing patches. On Google’s internal tests, Carbon produced secure code at more than twice the success rate of Argon. Because those abilities can be misused, Google first released the model only to selected defenders through its Fairwind Program while it tested safeguards. Security teams can still apply to that programme.

Gemini 4 Carbon Price and Access

Here is how access and pricing look after this week’s rollout, based on details reported by 9to5Google and Reuters.

Item Details
Consumer access Google AI Ultra subscribers, in the Gemini app, AI Mode and Antigravity
Developer access Gemini API paid tier, in AI Studio and Android Studio
Enterprise access Gemini for business, Antigravity and specialised agents
Introductory API price (till November 30) $2 per million input tokens, $10 per million output tokens
API price from December $4 per million input tokens, $20 per million output tokens
AI Pro users Continue to get Gemini 4 Argon
Scientists Can apply for free six month access to Argon and Carbon

The introductory price matches what Google charged for Argon at launch. Doubling the rate after November 30 means developers testing Carbon now pay half of what they will pay from December. Google has also removed the older Gemini 3.1 Pro from the Gemini app.

How the New Swarm Feature Works

The headline feature in this release is /swarm, now in preview inside Google Antigravity, the company’s tool for building with AI agents. According to Google, it “turns a single prompt into a coordinated team of agents” that users can track in real time.

In simple terms, instead of one AI working through a big job step by step, the system splits the job among several agents that work on different parts at the same time. A developer could, for example, ask for a full web app, and separate agents might handle the design, the code and the testing. Users can watch progress as it happens. Since it is still a preview, results and limits may change before a full launch.

Benchmark Claims and What They Mean

Google, together with independent analysis firm Artificial Analysis, introduced a new test called the Multimodal Knowledge Agents benchmark, which covers 200 tasks. Google says Carbon ranked first on it while using about three times fewer tokens than the next best model. Fewer tokens can mean lower cost and faster answers for the same task.

As with any company announced benchmark, these results come from Google and its partner. Independent testing by developers over the coming weeks will give a clearer picture of how Carbon compares with rival models in real use.

Key Highlights of the Gemini 4 Carbon Rollout

  • Wider rollout announced on November 10, 2026
  • Available to AI Ultra subscribers and paid Gemini API developers
  • Introductory price of $2 and $10 per million input and output tokens till November 30
  • Price doubles to $4 and $20 from December
  • New /swarm feature in Antigravity runs teams of agents from one prompt
  • Argon remains the default for AI Pro users and most coding work
  • First released on October 28 to cybersecurity defenders only

What It Means for Indian Developers and Startups

For Indian developers and startups that build products on top of large AI models, the main questions are access and cost. The reports reviewed did not give a country by country list of where Carbon is available, so Indian users should check the official Google announcement and their account settings. For teams paying in US dollars, the introductory pricing window until November 30 is the cheapest time to test whether Carbon suits their work. Developers can try the model through Google AI Studio on the paid tier.

The launch also comes during a busy period for the AI industry. Anthropic has set a price range for its stock market listing, which we covered in our report on the Anthropic IPO price band. With Carbon now widely available, competition among the leading AI labs for developers and enterprise customers is set to grow over the coming months.

More Technology News