Google announced Gemini 4 Carbon on November 10, describing it as its strongest model, with frontier performance in cybersecurity and long-horizon agentic workflows. Carbon follows the earlier Gemini 4 Argon release. Google says Argon remains the right choice for most software engineering and knowledge work, while Carbon is aimed at harder tasks that consume more tokens.
Carbon is available through Google AI Studio and Gemini for business for paying developers and enterprises, and is rolling out to Google AI Ultra subscribers in Google Antigravity, the Gemini app and AI Mode in Search. Cybersecurity access runs through Google's Fairwind Program on a rolling basis, and scientists can apply for six months of free access to Argon and Carbon at Antigravity's highest usage tier. Google is also previewing a /swarm feature in Antigravity that coordinates teams of agents.
Introductory pricing matches Argon at $2 per million input tokens and $10 per million output tokens. After the introductory period ends on November 30, 2026, the public Carbon model moves to $4 per million input tokens and $20 per million output tokens.
Google reports that 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, and second on the Multimodal Knowledge Agent benchmark. On safety, Google says Carbon blocked 99% of attack trajectories from an adaptive red-teaming agent in simulation, with the per-attempt breach rate falling to 0.8% from 3.1% on Argon, and that it passed 92% of a 375-task agentic honesty benchmark versus 70% for Argon. Google acknowledges that prompt injection robustness remains an area for continued improvement. These figures are reported by Google and its partners and have not been independently verified.
For developers, the main practical points are the pricing change on November 30, which affects cost estimates for any workload moved to Carbon now, and Google's own guidance that Argon stays the default for routine coding and knowledge tasks. Source: Google blog (blog.google).