Google has introduced Gemini 4 Carbon, describing it as its strongest model to date for cybersecurity, coding and long-running engineering work. The catch: it is only available to a small group of cybersecurity defenders for now.

What Carbon is built for

  • Announced on 28 October 2026 by Google DeepMind, Carbon builds on Gemini 4 Argon, which reached general availability on 13 October
  • It is designed to find software vulnerabilities, test whether attackers could exploit them, and write fixes
  • Google says Carbon more than doubles Argon’s success rate on its internal secure-code generation benchmarks
  • It also targets long-horizon agentic work — engineering tasks that need sustained effort over long periods

The benchmark scores

  • No. 1 on the Artificial Analysis Intelligence Index v5.0, with a score of 59
  • 88.2% on SEC-bench Pro for finding real-world software vulnerabilities
  • 54.8% on ExploitGym for validating vulnerabilities through exploitation
  • 71.0% on CWE-bench for generating working fixes for identified issues
  • These are Google’s own reported figures, not independently verified results

Who can use it — and what it costs

  • Access starts with a subset of cybersecurity defenders, plus the US government’s voluntary pre-release model testing programme
  • Access will gradually expand to trusted partners through Google’s Fairwind Program, with broader access for developers, businesses and consumers planned later — no public date has been announced
  • Introductory pricing matches Argon: $2 per million input tokens and $10 per million output tokens, later rising to $4 and $20 per million tokens

Related articles on HimachalGovt