Gemini 4 Carbon: Google’s Cybersecurity AI Model

Gemini 4 Carbon: Google Launches AI Model Focused on Cybersecurity

Google has unveiled Gemini 4 Carbon, a new AI model designed for cybersecurity work and complex, long-running coding tasks. However, access is limited at launch. Google says it will first share the model with selected cyber defenders and trusted partners while it continues to test safeguards.

What Gemini 4 Carbon is designed to do

Google says Gemini 4 Carbon can help security teams find software vulnerabilities, assess their risks and develop fixes. The company also describes it as a model for lengthy engineering tasks that may require sustained work across a codebase.

To support its claims, Google reported scores of 88.2% on SEC-bench Pro for vulnerability discovery, 54.8% on ExploitGym for testing whether vulnerabilities can be exploited, and 71% on CWE-bench for patching. These are figures published by Google. The company says the external benchmark evaluations were conducted without guardrails, so the scores should be understood in that context.

Google also reported stronger secure-code generation on its internal tests compared with Gemini 4 Argon. However, internal results are not the same as independent evaluations. The launch announcement does not establish how the model performs across every real-world cybersecurity setting.

Gemini 4 Carbon access will begin with selected defenders

Google says it began offering Gemini 4 Carbon to a subset of cyber defenders during the week of its October 28 announcement. It plans to extend access to more trusted partners through its Fairwind Program.

For now, developers, businesses and consumers do not have general access. Google says it will gather feedback and continue refining safeguards before making the model available more broadly through its products. Reuters also reported that Google is initially limiting access to selected cybersecurity users.

Google announces introductory pricing

Google announced introductory prices of $2 per million input tokens and $10 per million output tokens. After the introductory period, the company says those rates will rise to $4 and $20, respectively. These are announced prices; they do not mean the model is currently open to all API users.

The staged rollout reflects the model’s cybersecurity capabilities and the need to assess them carefully. As a result, wider availability will depend on Google’s testing and deployment plans. The company has not announced a date for general access.

The Nation report describes Google’s unveiling of Gemini 4 Carbon. Google’s announcement provides the model’s stated capabilities, benchmark figures, pricing and initial access plan. Those details remain company-reported claims unless independently assessed.

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top