AI

Google DeepMind Launches Gemini 4 Carbon for Cybersecurity and Long-Horizon Coding

Wednesday, October 28, 2026Read Original

Details

  • Google DeepMind announced Gemini 4 Carbon on October 28, 2026, describing it as its next frontier model.
  • The model targets cybersecurity defense, coding, and agentic tasks that may require sustained work over multiple days.
  • Carbon is designed to detect vulnerabilities, assess exploitability, and help develop remediation patches across the cybersecurity defense lifecycle.
  • Google reported scores of 88.2% on SEC-bench Pro, 54.8% on ExploitGym, and 71.0% on CWE-bench.
  • An initial group of cyber defenders will test Carbon before Google expands access to additional trusted partners through its Fairwind Program.
  • Google plans to make the model available more broadly to developers, businesses, and consumers, but did not announce a public release date.
  • The limited rollout gives Google time to refine safeguards and guardrails before wider deployment of a model capable of sustained autonomous coding and security work.

Impact

Carbon places Google in closer competition with frontier-model providers pursuing agentic coding and cybersecurity applications. Its staged release acknowledges that long-running autonomous systems create higher safety and misuse risks than conventional chatbots. If the reported benchmark performance translates into production environments, Carbon could improve vulnerability discovery and patch development while increasing pressure on rivals to offer similarly specialized models and stronger controls.

Rift Dispatch
Google DeepMind Launches Gemini 4 Carbon for Cybersecurity and Long-Horizon Coding | RiftLab