Google Launches Gemini 4 Carbon, Its 'Strongest Model' Yet, With Swarm Tool That Turns a Prompt Into a Team

MOUNTAIN VIEW, Calif. — Google on Tuesday released Gemini 4 Carbon, which it calls its most powerful artificial intelligence model, to developers, businesses and paying consumers. It also previewed a "swarm" feature that lets many AI agents work together on large projects.

Google said Carbon is its "strongest model" yet and can compete with the best systems from Anthropic and OpenAI, The Verge reported. The release follows a two-week preview limited to cybersecurity professionals through Google's Fairwind Program.

Built for the hardest tasks

Carbon joins Gemini 4 Argon, the frontier model Google released last month. The company said Carbon "pushes the frontier even further," though Argon will likely remain "the ideal choice for the majority of software engineering and knowledge work."

Carbon, Google said, is "built for tasks that test the limits of a model's intelligence," including complex multi-step coding, open-ended scientific research and long-running AI agent work that requires a lot of reasoning and checking.

Google is emphasizing cybersecurity. The company said its Cloud Vulnerability Research team used Carbon to show that two bugs in open-source software could be exploited, leading to critical classifications and patches after responsible disclosure.

Swarm: one prompt, many agents

The most notable addition is a preview of a feature called /swarm in Google Antigravity, the company's platform for agent-based software development. Google said the tool turns a single prompt "into a coordinated team of agents" that users can monitor in real time.

Users can assign Carbon to high-level tasks such as architecture and code review while other agents build and test changes in parallel. Google said it tested /swarm internally for five weeks before the preview.

In one demonstration, a team of Carbon agents turned historical photos and engineering drawings into an interactive 3D exhibition of the Eiffel Tower's construction from one prompt.

The /swarm preview is available in the Antigravity app and command-line tool.

Where to get it

Carbon is available through the paid tier of the Gemini API in Google AI Studio and Android Studio, and through Gemini for business for paying developers and companies. It is also rolling out to Google AI Ultra subscribers in the Gemini app, AI Mode in Google Search and Antigravity.

Developers can call it as "gemini-4-carbon" with a context window of 1 million tokens, a measure of how much text the model can process at once.

GitHub said the same day that Gemini 4 Carbon is generally available in GitHub Copilot for Pro+, Max, Business and Enterprise subscribers. GitHub said the model is designed for complex coding, agent and reasoning tasks, with a focus on secure code generation. Users can select it in Visual Studio Code, JetBrains tools, Xcode, the Copilot command-line tool and GitHub's mobile apps, among others.

Pricing

Google is offering Carbon at an introductory price of $2 per million input tokens and $10 per million output tokens, the same as Argon. Prices are set to double to $4 and $20 at the end of November.

Even at the higher rates, Carbon remains cheaper than OpenAI's GPT-6 Astra and Anthropic's Fable 5.2, according to reports comparing the companies' published pricing.

A crowded race

The launch is Google's latest move in a fierce contest with OpenAI, Anthropic and others to build the most capable AI systems, especially for coding and autonomous agents, where companies see the biggest near-term business opportunities.

Multi-agent systems, in which several AI models split up and coordinate work, have become a major focus across the industry as companies try to move AI from answering questions to finishing long, complex projects with little human help.