Google Launches Gemini 4 Carbon AI Model to Boost Cybersecurity and Software Defense

Google launched its newest frontier artificial intelligence model named Gemini 4 Carbon on Wednesday, October 28, to help find and fix security vulnerabilities. ## Capabilities And Performance This release follows the rollout of Gemini 4 Argon just two weeks prior. Google reports that Carbon outperforms Argon in coding tasks, finding software bugs, generating secure code, and handling agentic work. It can run for days to finish long engineering jobs without taking shortcuts and resists prompt injection attacks. It also cuts unsupported claims by half compared to Argon and produces vulnerability-free code twice as often. Google Chrome engineers found that Carbon generates high-quality patches ten times better than non-Google models. On benchmarks, Carbon ranks number one on the v5.0 Artificial Analysis Intelligence Index. It also scores 88.2 percent on SEC-bench Pro, 54.8 percent on ExploitGym, and 71.0 percent on CWE-bench. It beat rival models including OpenAI's GPT-6 Astra, Anthropic's Claude Opus 5.5, and Claude Fable 5.1 in tests. ## Pricing And Availability Google is using a careful rollout approach by giving access first to a small group of trusted cybersecurity defense companies starting this week. The model will later expand to more partners through the Fairwind Program before a wider public release. At launch, Carbon costs two dollars per million input tokens and ten dollars per million output tokens, matching the price of Argon. The launch arrived alongside Google's earnings report showing strong cloud computing growth but weak search advertising revenue, causing company stock to drop nearly ten percent in after-hours trading. ## Context Google releases major artificial intelligence models periodically to compete with rivals like OpenAI and Anthropic in the fast-moving tech market. Cybersecurity testing benchmarks help developers measure how well computer models find real-world software bugs and create safe patches. Gradual rollouts let tech companies test new tools with select security experts before making them available to the general public.