Google announces Gemini 4 Carbon with a limited rollout for cyber defenders
Google says its new AI model can find vulnerabilities, develop exploits and patch code, but its reported performance figures come from benchmarks and company testing.
Google announced Gemini 4 Carbon on 28 October 2026, with initial access for a subset of cybersecurity defenders, according to Reuters. The model targets vulnerability discovery, exploitation and software patching, tasks that could help security teams find and fix flaws but also require careful controls over access.
Reuters reported that access will begin through Google's Fairwind Program while the company continues testing safeguards. Google says it intends to bring Carbon to developers, businesses and consumers later, but has not given a date for wider availability.
What Google says Carbon can do
Google calls Carbon its strongest model and says it is built for cybersecurity and long-running engineering work. It reports scores of 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation and 71.0% on CWE-bench for patching. These are benchmark results reported by the company; they do not establish how reliably the model will perform in every live security setting.
The company says its teams have used Carbon to trace and patch subtle bugs across large codebases and low-level systems, sometimes working through multi-step tasks over days. It also says Carbon ranked first on version 5.0 of the Artificial Analysis Intelligence Index with a score of 59.
How the rollout compares with Gemini 4 Argon
Carbon follows Google's earlier Gemini 4 Argon announcement by two weeks, Reuters reported. Google says Carbon generated secure code at more than twice Argon's success rate in internal benchmarks. That comparison reflects Google's tests, rather than an independently established improvement for every coding task.
Reuters reported introductory prices of $2 per million input tokens and $10 per million output tokens, matching Argon's introductory prices. Those figures matter to developers weighing the cost of using Carbon for extended coding or security tasks once they can obtain access.
The wider industry is also focusing on AI-related security threats. At an Oracle AI World event in Las Vegas, OpenAI chief executive Sam Altman urged companies to understand and mitigate those threats, The National reported. His remarks illustrate the attention security is receiving across the sector; they were not an assessment of Carbon.
Read in depth: Technology
Amazon’s new Echo Dot is on sale, while Echo Show 16 awaits December launch
The Echo Dot adds smart home hub and eero extender functions. The larger Echo Show 16 can be preordered, with The Verge reporting a December 16 launch.
South Korea fines Google and Apple over app store payment rules
The regulator imposed 63.71 billion won in fines and ordered changes to payment options for app developers, with compliance deadlines tied to the companies’ receipt of...
Sources and context
- Google launches Gemini 4 Carbon AI model with cybersecurity focusReuters, republished by CNA
- Gemini 4 Carbon: our next frontier modelGoogle
- OpenAI’s Sam Altman urges companies to focus on AI security threat as GPT-6 adoption soarsThe National
AI-assisted article checked against the listed sources. NewsJaws did not conduct interviews or attend the reported events.