Google Launches Gemini 4 Carbon AI Model With Cybersecurity Focus

Google Announces Gemini 4 Carbon

Google has officially revealed gemini 4 carbon, its strongest artificial intelligence model to date, featuring specialized upgrades for cybersecurity defense and long-horizon tasks. The new system succeeds Gemini 4 Argon, which was introduced two weeks prior.

Focus on Cybersecurity and Long-Horizon Tasks

Designed to tackle complex and multi-step engineering challenges, the model can run for days to persist through dead ends. In testing environments, the software demonstrated substantial improvements across coding, agentic workflows, and web development.

The system delivers prominent performance across the entire cybersecurity defense lifecycle, covering vulnerability detection, risk validation through exploitation, and root-cause remediation. External evaluations without guardrails show strong scores, including top placement on SEC-bench Pro and ExploitGym.

Safety, Rollout, and Pricing

Google is taking a deliberate approach to deployment, engaging with the United States government's voluntary pre-release testing process. Access begins with select cyber defenders and will expand via the Fairwind Program before rolling out broadly to consumers and developers.

Pricing matches the previous Argon tier, set at $2 per million input tokens and $10 per million output tokens, with cached inputs discounted by 95 percent. Safety measures include enhanced alignment training to reduce misalignments and curb unsupported claims.

Sources: blog.google, Reuters, 9to5Google.