Google Unveils Gemini 4 Carbon AI Model With a Focus on Cybersecurity
Google announced its newest AI model, Gemini 4 Carbon, on Wednesday, citing gains in coding and cyberdefense. For now, it is available only to a select group of cybersecurity companies.
Technology company Google on Wednesday announced Gemini 4 Carbon, its latest frontier artificial intelligence (AI) model, saying it brings significant gains in coding, agentic tasks (where AI carries out work on its own) and cybersecurity. The release matters because AI models are increasingly used to find software vulnerabilities and to produce safer code.
According to Google, Carbon holds the top spot on the Artificial Analysis Intelligence Index (v5.0) benchmark. Claude Opus 5.5 and GPT-6 Astra rank behind it.
Improvements over Argon
Carbon builds on Argon, a model Google released earlier this month. The company describes what it calls a “significant leap” in cyberdefense, saying the new model is better at finding vulnerabilities and generating secure code. Carbon currently ranks first on SEC-bench Pro, a benchmark measuring an AI model’s ability to discover real-world bugs, and on ExploitGym, which measures how well a model can turn a vulnerability into an exploit.
Google says the model is designed for longer tasks. Its internal teams reportedly saw it run for days, persisting through dead ends to handle complex, multi-step engineering work. The company adds that Carbon is better at avoiding shortcuts, resisting prompt injection attacks and collaborating with users. It is also meant to be more transparent and honest: Google says it cuts unsupported claims by half compared with Argon.
Availability and pricing
Carbon is currently available only to a select group of cybersecurity defense companies. Google plans to expand access to other partners in its Fairwind Program over time. The model launches at the same introductory price as Argon: $2 per million input tokens and $10 per million output tokens.
