Google has launched Gemini 4 Carbon, its latest frontier model, which demonstrates significant advancements in cybersecurity, coding, and agentic tasks. The model tops the Artificial Analysis Intelligence Index (v5.0) benchmark, surpassing competitors like Claude Opus and GPT-6 Astra. This marks a major step forward in AI-driven security solutions, with Google emphasizing its ability to detect vulnerabilities and generate secure code more effectively than previous versions.
Currently available to select cybersecurity defense companies, Gemini 4 Carbon is part of Google’s Fairwind Program and will expand to more partners in the future. The model's performance on SEC-bench Pro and ExploitGym highlights its capability to identify real-world bugs and convert vulnerabilities into exploit scenarios. Google claims the model is designed for complex, multi-step tasks, with internal teams reporting it can operate for extended periods to overcome challenges.
Gemini 4 Carbon also improves on transparency and honesty, reducing unsupported claims by two times compared to its predecessor, Argon. The model maintains the same pricing structure as Argon, at $2 per million input tokens and $10 per million output tokens. Google highlights its resistance to prompt injection attacks and improved collaboration with users, making it a robust tool for advanced cybersecurity applications.
The release of Gemini 4 Carbon underscores Google’s commitment to pushing the boundaries of AI in security. With its focus on long-term task execution and enhanced security features, the model is positioned to become a key player in the AI-driven cybersecurity landscape. As it expands to more partners, it could significantly influence the industry’s approach to threat detection and response.