Google has officially begun the general rollout of its most powerful artificial intelligence model — Gemini 4 Carbon. This happened two weeks after its limited release for cybersecurity specialists.

The tech giant claims that Carbon is the "most powerful model," capable of competing on equal terms with flagship solutions from Anthropic and OpenAI.

Specialization in complex and long-running tasks

The release of Carbon came just a month after the launch of another expanded model from the company — Gemini 4 Argon. According to Google representatives, Argon will remain the optimal choice for most standard software development and analytical tasks.

Instead, Gemini 4 Carbon was designed for tasks that require the utmost level of AI intelligence:

complex multi-step programming over a long time horizon;

New Swarm feature: a team of agents from a single prompt

To demonstrate Carbon's capabilities in executing lengthy and large-scale projects, Google is launching a preview of the Swarm feature on its Antigravity development platform, writes The Verge.

This tool allows turning a single text prompt into a coordinated group of AI agents that distribute roles among themselves and work as a single team to solve a complex task.

Access conditions and price dumping

for developers and companies via the Gemini API in Google AI Studio and the Gemini for Business service;

for private users in the Gemini app for Google AI Ultra tariff subscribers, in AI Mode in Search, and on the Antigravity platform.

The company has not yet specified whether the model will be available to users of basic tariffs and the free version.

As with Argon, Google is entering the market with aggressive dumping. The starting price is $2 per 1 million input tokens and $10 per 1 million output tokens. This is approximately five times cheaper than OpenAI GPT-6 Astra and four times cheaper than Anthropic's Fable 5.2 model.

Even after the promotional period ends at the end of November, when prices double ($4 / $20), Carbon will remain significantly more cost-effective than competitors per token.

Benchmarks and performance

Google's published tests show that Carbon holds leading positions in the market across most metrics, trailing only the newest Fable 5.2. However, experts traditionally urge caution in interpreting such comparisons, as tech giants often select only the most favorable benchmarks for presentations.