Google said on Nov. 10 that Gemini 4 Carbon is now available1 to paying developers, enterprises, Google AI Ultra subscribers and participants in its scientist access program, according to the DeepMind Blog1.
The release matters because Google is positioning Carbon as a higher-end model than Gemini 4 Argon for long-horizon coding, knowledge work, scientific reasoning and cybersecurity, while charging the same introductory API price of $2 per million input tokens and $10 per million output tokens, according to the DeepMind Blog1.
Automatica reported earlier on Nov. 10 that a social-media claim about Gemini 4 Carbon lacked corroboration at the time, and the company's post now provides the first direct public documentation in the available evidence for the model's existence and rollout Automatica Press4. Google said Carbon follows Gemini 4 Argon and that Argon remains its default choice for most software engineering and knowledge-work tasks, while Carbon is aimed at more demanding multi-step coding, debugging and research workloads DeepMind Blog1.
Google's announcement cites benchmark and evaluation results but they are largely vendor-supplied. The company said Carbon ranks No. 1 on CursorBench 4.0, Perplexity's WANDR and Zapier's AutomationBench, and No. 2 on the new Multimodal Knowledge Agent benchmark, while claiming lower cost per task than competing models DeepMind Blog1. Perplexity's WANDR benchmark post2 describes a 500-task evaluation for research agents, but the excerpt in the available evidence reports results for "Perplexity Search as Code" and does not independently verify Google's Carbon ranking.
Google also said Carbon had been released two weeks earlier to cybersecurity defenders through its Fairwind Program DeepMind Blog1. The current Fairwind Program page3 in the available evidence describes early-access cyber defense access for Gemini 4 Argon and says the program works with more than 650 partners globally.