Google made Gemini 4 Carbon broadly available on 10 November, opening its strongest model yet to paying developers, enterprises, scientists and consumers. Carbon is available through the Gemini API in Google AI Studio and Gemini for business, and Google AI Ultra subscribers get it in the Gemini app, AI Mode in Search and the agent-first platform Google Antigravity.

Google positions Carbon as its frontier model for cybersecurity and long-horizon agentic work across complex knowledge work, software engineering and scientific reasoning, while Gemini 4 Argon stays the recommended default for routine engineering and knowledge work. Introductory pricing matches Argon at $2 per million input tokens and $10 per million output tokens through 30 November, doubling to $4 and $20 from 1 December. The free tier of the Gemini API does not include Carbon.

Two weeks before the public rollout, Carbon went to a small group of cybersecurity defenders through Google's Fairwind Program. On Google's own benchmarks it scores 53.9% on AutomationBench and 88.2% on SEC-bench Pro for finding vulnerabilities, with a 0.0% attack success rate on the Gray Swan prompt-injection benchmark against 0.7% for Argon. Those figures are company-reported, not independently verified.

Antigravity is also previewing /swarm alongside Carbon: a coordinated multi-agent feature that turns one prompt into a tracked team of agents splitting roles such as design, verification and auditing for long-horizon workflows. Carbon is the second Gemini 4 frontier model after Argon, which Google announced on 28 October.

Sources