Google is making Gemini 4 Carbon available to the general public, two weeks after giving a select group of cybersecurity defenders early access to the model. The company says Carbon is its strongest model to date and can hold its own against the best systems from Anthropic and OpenAI. The wider rollout puts the model in front of ordinary users and developers in the US and elsewhere.
According to Google, Carbon extends the frontier beyond Argon, the company's other frontier model, which launched just last month. Google said Argon will likely stay the best fit for most software engineering and knowledge work, while Carbon is designed for tasks that stretch a model's intelligence, including complex multi-step coding and open-ended scientific research.
Google is also previewing a swarm feature on Antigravity, its agentic development platform. The feature is meant to let users turn a single prompt into a coordinated team of agents, which Google presented as a sign of Carbon's value for long-horizon work involving AI agents.
Pricing follows the pattern set by Argon. Carbon launches at an introductory $2 per million input tokens and $10 per million output tokens, roughly a fifth of the cost of OpenAI's GPT-6 Astra and a quarter of the cost of Anthropic's Fable 5.2, which launched yesterday. Those rates double at the end of November when the introductory period ends, though Google says Carbon will remain significantly cheaper per token than frontier rivals even then. Per-token savings do not automatically mean lower costs overall, since models can consume different numbers of tokens on the same tasks.
Benchmarks Google cited when it first announced Carbon place the model among the most powerful on the market on most metrics, behind Fable 5.2. Such figures come with a caveat: companies can and do cherry-pick benchmark results to present their tools in the best light.
The release, arriving quickly after Argon, marks a decisive return for Google to the very forefront of AI research. For months, frontier development had looked like a two-horse race between Anthropic and OpenAI.
More AI news from TechManNews.
