Google announced Gemini 4 Carbon on October 28, 2026, describing it as a new frontier model focused on cybersecurity and long-horizon agentic work. Google reported scores of 88.2% on SEC-bench Pro for vulnerability discovery, 54.8% on ExploitGym for exploitation, and 71.0% on CWE-bench for patching. The company also said Carbon improves coding and agentic performance and ranks first on the v5.0 Artificial Analysis Intelligence Index. Initial access is limited to a subset of cyber defenders through Google’s Fairwind Program, with broader access for developers, businesses, and consumers planned later.
Why it matters
The reported results could matter to developers and security teams evaluating AI-assisted vulnerability discovery, exploit validation, and code repair. However, the SEC-bench Pro and ExploitGym figures are Google-reported evaluations, and public benchmark pages do not yet independently list Carbon for comparison.
Key points
- Google announced Gemini 4 Carbon on October 28, 2026, as a frontier model for cybersecurity and long-horizon agentic tasks.
- Google reported 88.2% on SEC-bench Pro for finding vulnerabilities and 54.8% on ExploitGym for exploitation.
- Google also reported 71.0% on CWE-bench for patching vulnerabilities.
- The company described gains in coding and agentic tasks, but the cited cybersecurity scores are company-reported rather than independently verified on the public benchmark leaderboards.
- Access initially goes to selected cyber defenders through the Fairwind Program; broader availability is planned, but no general release is open yet.
Based on reporting by The original announcement and The Hacker News.
Join the conversation! Your thoughts help the community grow.