Gemini 4 Carbon: our next frontier model

Why it matters

Google has announced Gemini 4 Carbon, describing it as its next frontier model with a particular emphasis on cybersecurity and long-horizon engineering work. The model tops the latest Artificial Analysis Intelligence Index and posts leading results on benchmarks covering vulnerability discovery, exploitation, and patching, building on Gemini 4 Argon, which launched shortly before. The announcement is notable for how the company frames safety and reliability alongside capability: it reports reduced cheating on coding evaluations, fewer misalignment alerts in live internal use, fewer unsupported claims, and better behavior when facing ambiguous instructions, such as surfacing discrepancies rather than guessing. The rollout is deliberately staged. Access begins with a subset of cyber defenders and expands gradually to trusted partners through the Fairwind Program, while the company participates in the U.S. government's voluntary pre-release testing process. Wider availability through consumer and developer products will follow once guardrails are refined. The pricing matches Argon's introductory rates. For readers, the story matters because it shows a frontier lab coupling aggressive capability gains with a cautious, security-focused deployment strategy.