Google rolls out Gemini 4 Carbon with agent swarm for developers

Google has opened Gemini 4 Carbon to all paying developers, adding a “swarm” capability that lets a single prompt spawn a coordinated team of agents. The rollout also brings a steeply discounted token price that could make frontier‑level AI affordable for small firms.

What happened

According to The Verge, Google announced today that Gemini 4 Carbon is now available via the Gemini API in Google AI Studio, Gemini for Business, and to all Google AI Ultra subscribers in the Gemini app, Search’s AI Mode, and the Antigravity platform. The model is billed as Google’s “strongest” yet, built for “tasks that test the limits of a model’s intelligence,” such as multi‑step coding and open‑ended scientific research.

The Verge notes that Carbon “pushes the frontier even further” than the Argon model released a month earlier, while Argon remains the “ideal choice for the majority of software engineering and knowledge work.”

A key highlight is the preview of a “swarm” feature on Google’s agentic development platform Antigravity. The feature promises to turn a single prompt into a coordinated team of agents, aiming to simplify long‑horizon, multi‑agent workflows.

Pricing is aggressively low: an introductory $2 per million input tokens and $10 per million output tokens—about one‑fifth the cost of OpenAI’s GPT‑6 Astra and one‑quarter the cost of Anthropic’s Fable 5.2. After the introductory period ends in late November, the rates double to $4 and $20 per million tokens, but remain cheaper on a per‑token basis than the rivals.

Benchmarks cited by Google place Carbon among the most powerful models on the market, trailing only Anthropic’s Fable 5.2. The Verge cautions that benchmark cherry‑picking is common, so the exact performance edge may vary by use case.

Why it matters for small‑business owners and team leads

First, the price point directly attacks the biggest barrier for SMEs: cost. At $2/$10 per million tokens, a typical weekly workflow—say 10 k input tokens and 5 k output tokens—would cost under $0.10 during the intro period, a fraction of what a comparable OpenAI or Anthropic call would charge. Even after the November price bump, the expense stays modest enough to experiment without draining budgets.

Second, the swarm capability could democratize multi‑agent automation. Small teams often lack the engineering bandwidth to stitch together several specialized agents. By letting a single prompt orchestrate a “team of agents,” Antigravity lowers the technical threshold for building complex pipelines such as automated lead qualification, multi‑step data enrichment, or iterative product‑design simulations. In practice, this means a sales lead can be scored, enriched, and routed—all without writing custom orchestration code.

Finally, Gemini 4 Carbon’s focus on “complex, multi‑step coding and open‑ended scientific research” signals that Google is positioning the model for high‑cognitive tasks that previously required multiple specialized tools. For a boutique consultancy, this could translate into faster prototype generation, deeper market‑analysis models, or more nuanced natural‑language reports, all within a single API.

What to do this week

1. Create a Gemini API key in Google AI Studio and run a quick token‑cost test on a real‑world prompt to see how the $2/$10 rates compare to your current provider.
2. Spin up a sandbox project on Antigravity and try the swarm preview by asking the platform to “research competitor pricing, summarize findings, and draft a one‑page brief.”
3. Map one repetitive workflow (e.g., lead enrichment) to a multi‑agent design and prototype it with Carbon’s API, measuring time saved versus your existing manual process.

Sources

• Google releases Gemini 4 Carbon with agent ‘swarm’ feature — The Verge - AI