Google DeepMind has announced Gemini 4 Carbon, its newest frontier model, which the company says is its strongest yet, with state-of-the-art cybersecurity and long-horizon capabilities. Carbon ranks first on the updated v5.0 Artificial Analysis Intelligence Index with a score of 59. Google says internal teams have seen it run for days, persisting through dead ends to root-cause and patch subtle bugs across large codebases. It also gained 89 Elo points on WebDev Arena, reaching fourth place with a score of 1768. Carbon follows Gemini 4 Argon, released two weeks earlier, and launches at the same introductory price of $2 per million input tokens and $10 per million output tokens, with cached input tokens priced 95% below standard input. The cybersecurity results are the centrepiece. Carbon scores 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation and 71.0% on CWE-bench for patching, which Google describes as state-of-the-art and measured without guardrails. Google also says the Chrome team found Carbon produced high-quality patches more than 10 times better than those from another lab's large frontier model, and that Wiz sees it leading on web penetration testing benchmarks. On safety, Google reports that successful cheating on Terminal-Bench 4.0 fell to 0.0%, high-severity misalignment alerts in internal live coding environments dropped by almost 80%, and unsupported claims were roughly halved compared with Argon. Access is deliberately restricted. From this week, Carbon is available to a subset of cyber defenders, with other trusted partners to be added gradually through the Fairwind Program. Google says it is taking part in the U.S. government's voluntary pre-release testing process and will expand access to developers, businesses and consumers as soon as guardrails are ready, but it has not given a date. For Australian businesses and professionals, nothing changes immediately. Carbon is not generally available, and most local organisations cannot use it until it reaches broader release. When it does, the introductory token pricing will matter most for teams building agentic coding tools, and any personal information sent to the model remains subject to the Privacy Act 1988, including its rules on overseas disclosure. Several of the headline figures are Google's own internal or vendor-reported results, so businesses should wait for independent testing before relying on them. Security teams already working with Google's partner programs may want to watch the rollout closely.