# Google releases Gemini 4 Carbon, its strongest model

> Google released Gemini 4 Carbon via API and to Ultra subscribers.

*Carbon targets agentic work and cybersecurity, and Google is previewing a multi-agent /swarm mode alongside it.*

By Bez · FeaturedDaily
Canonical: https://featureddaily.com/news/google-releases-gemini-4-carbon-its-strongest-model

**What happened:** Google released Gemini 4 Carbon on 10 November and called it the company's strongest model. It is available through Google AI Studio and Gemini for business. It is also rolling out to Google AI Ultra subscribers in Google Antigravity, the Gemini app and AI Mode in Search. [Google's announcement](https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-carbon-now-available/) has the details.

**The numbers:** The introductory API price is $2 per million input tokens and $10 per million output tokens, the same as Argon. From 30 November, the price rises to $4 input and $20 output per million tokens. Google says 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. It ranks second on the Multimodal Knowledge Agent benchmark, at more than 6X lower average cost per task than the top model. These figures are Google's own and have not been independently verified.

**Why it matters:** Google built Carbon for long-horizon agentic workflows, complex knowledge work, software engineering and scientific reasoning. It also targets cybersecurity, where Google says Carbon leads on finding, exploiting and patching vulnerabilities in SEC-bench Pro, ExploitGym and CWE-bench v1. Defenders can get access through the Fairwind Program. Google reports that Carbon blocked 99% of attack trajectories in simulated tests, and that its per-attempt breach rate fell to 0.8% from 3.1% on Argon.

**The catch:** Google recommends Argon for most software engineering and knowledge work. Carbon uses more tokens to reason deeply and run extended verification loops.

**In their words:** Koray Kavukcuoglu, SVP at Google DeepMind and Chief AI Architect at Google, wrote in the announcement that Carbon is "our strongest model with frontier performance in cybersecurity capabilities."

**What's next:** Google is previewing /swarm in Google Antigravity. It turns one prompt into a coordinated team of agents, tracked in real time. Carbon handles high-reasoning roles such as architecture, coordination and auditing, while faster models run in parallel isolated workspaces, with adversarial verifiers checking every change. Scientists can apply for free six-month access to Argon and Carbon in Antigravity's highest usage tier, subject to eligibility and capacity.

## Key takeaways

- Introductory price: $2 input, $10 output per million tokens.
- Price rises to $4 input and $20 output from 30 November.
- Google says Carbon leads CursorBench 4.0, WANDR and AutomationBench.

## Sources

- [Gemini 4 Carbon now available](https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-carbon-now-available/) — Google, 2026-11-10
