The advanced AI model is designed for complex software engineering, scientific research and long-horizon agentic workflows, with enhanced cybersecurity capabilities.
Google has made Gemini 4 Carbon available to developers, enterprises, scientists and consumers, introducing its most advanced AI model for complex reasoning, cybersecurity and long-horizon agentic workflows.
According to Google, Gemini 4 Carbon delivers frontier-level performance across complex knowledge work, software engineering and scientific reasoning. The model is designed for tasks that require deeper reasoning, multi-step coding, debugging and open-ended scientific research, using more tokens than Gemini 4 Argon to explore problems, verify results and execute extended workflows.
The launch follows Google's earlier limited release of Carbon to cybersecurity defenders through its Fairwind Programme. The company said the model builds on the capabilities of Gemini 4 Argon, while targeting more demanding tasks that require advanced reasoning and extended verification.
Google highlighted Carbon's performance across independent benchmarks designed to evaluate AI models in practical workflows. The model secured the top position on CursorBench 4.0, Perplexity's WANDR and Zapier's AutomationBench. According to the company, it also delivers more than twice the cost efficiency per task compared with the next-best external model on these benchmarks.
Alongside the model's wider availability, Google is previewing the /swarm feature in Google Antigravity to support larger and more complex development projects.
The feature enables users to turn a single prompt into a coordinated team of AI agents that can be monitored in real time. Developers can assign Gemini 4 Carbon to specialised roles requiring deeper reasoning, including architectural design, coordination and auditing, while deploying faster, lighter models across parallel workspaces.
The system also incorporates adversarial verifiers to assess and test changes made during the development process, helping teams manage complex projects through coordinated AI workflows.
Gemini 4 Carbon is available through the API in Google AI Studio and Gemini for business for paying developers and enterprise customers. It is being offered at an introductory price of $2 per million input tokens and $10 per million output tokens, matching the introductory pricing of Gemini 4 Argon.
The model is also rolling out to Google AI Ultra subscribers through Google Antigravity, the Gemini app and AI Mode in Google Search.
With Gemini 4 Carbon, Google is expanding its AI model portfolio towards more sophisticated agentic applications, positioning advanced reasoning, multi-agent coordination and extended verification as key capabilities for complex enterprise and scientific workflows.
