MONDAY, NOVEMBER 16, 2026 Tech news, minus the noise.
N News Cast
LATEST
Home  /  AI
AI

Google Launches Gemini 4 Carbon, Its Strongest Model Yet for Cybersecurity and Long Coding Jobs

Google put its new frontier model Gemini 4 Carbon on general availability on November 10, pitching it at cybersecurity defenders and developers running long, complex jobs.

Google Launches Gemini 4 Carbon, Its Strongest Model Yet for Cybersecurity and Long Coding Jobs
Source: Asoundd, CC BY-SA 4.0, via Wikimedia Commons

Gemini 4 Carbon Is Here, and Google Is Aiming It at Serious Work

Google made Gemini 4 Carbon available on November 10, calling it the strongest model in its Gemini 4 family. The launch came two weeks after Google released Carbon to a group of cybersecurity defenders through its Fairwind Program, and it follows the release of Gemini 4 Argon, the model that opened the family. Carbon is now available through the API in Google AI Studio and Gemini for business, and it is rolling out to Google AI Ultra subscribers in Antigravity, the Gemini app, and AI Mode in Search.

The pitch is specific. Google describes Carbon as a frontier model for cybersecurity capabilities and long-horizon agentic workflows across complex knowledge work, software engineering, and scientific reasoning. That is a narrower claim than the usual "best at everything" launch language, and it tells you who Google thinks should pay attention first: security teams, developers running agents that work for hours rather than minutes, and researchers doing heavy analysis.

What Gemini 4 Carbon Costs

Pricing is straightforward, and it matches Argon. Carbon costs $2 per million input tokens and $10 per million output tokens at the introductory price. Google has not said when the introductory period ends, so treat that price as a starting point rather than a promise.

For teams already using Argon, the price parity matters. It means trying Carbon costs nothing extra on paper, and the decision becomes purely about whether the stronger model is worth the same money for your workload. Early independent rankings give a hint: one evaluation index placed Carbon third overall while also ranking it the fastest and cheapest of the top five models it tracks. Speed and price are not the same as quality, but for agentic work that runs many steps, both add up fast.

Why Google Led With Cybersecurity

The Fairwind Program detail is the most interesting part of the launch. Google gave Carbon to cybersecurity defenders two weeks before the public release. That is a deliberate choice. Security work is one of the few places where a stronger model has an obvious, measurable payoff: finding a vulnerability before an attacker does, or triaging an incident faster than a human team working alone.

It is also a field where the risks of a powerful model are obvious, which is probably why Google started with defenders rather than the general public. A model that is good at finding vulnerabilities is also good at finding them for the wrong people. Google has not published a detailed safety report for Carbon alongside the launch, and that is worth watching. The company has talked up frontier cybersecurity capability without saying much in public yet about the guardrails around it.

How Carbon Fits the Gemini 4 Family

Argon was the opening act: a general frontier model for everyday use across Google products. Carbon is the specialist sibling, aimed at the hardest, longest jobs. The pattern is familiar from other labs. A family of models with different strengths lets a company serve different customers without building one model that tries to do everything.

The timing also fits a busy month for model releases. November has already seen OpenAI launch GPT-6 Cyber, a security-focused model available only to approved organizations, and Moonshot put the full weights of Kimi K3.1 Pro on Hugging Face. Google is not launching in a vacuum. Carbon is its answer to the question every lab is asking right now: what is the next model actually for?

For most users, the honest answer is that Carbon changes little today. If you use the Gemini app or AI Mode in Search, you will get it as part of your subscription. If you build with the API, it is another option at the same price as Argon. The people who should look closely are the ones running long agent workflows, security teams evaluating AI help, and developers who have been waiting for a model that can hold a complex task together over many steps. For them, the launch is worth a test run this week, while the price is still at its introductory level.

The Bottom Line on Gemini 4 Carbon

Carbon is Google's strongest model, aimed at cybersecurity and long-horizon work, priced the same as Argon at $2 in and $10 out per million tokens. It is available now through the API and rolling out to Ultra subscribers. The launch is solid but not dramatic. The real test will be whether teams doing serious security and engineering work find that Carbon holds up over long jobs where earlier models drifted, repeated themselves, or lost track. That verdict will take weeks of real use, not a launch blog post. If you want a broader view of how the model race is shaping up, read our Gemini 4 Argon release coverage and our look at the AI agent wars, where the fight over who builds the best working agent is already underway. And if you are wondering how these models fit into everyday tools, our GPT-6 rollout piece covers the other side of the same race.

AS
News Cast Editorial Team

We cover AI, gadgets, startups and security — original reporting and explainers, written by humans, for humans. Got a tip? Reach us via the contact page.