Google unveils Gemini 4 Carbon, its "strongest model with state-of-the-art cybersecurity" capabilities, ranked #1 on Artificial Analysis' new Intelligence Index

AI models can now find and patch software vulnerabilities at a high success rate.

Google DeepMind has unveiled Gemini 4 Carbon, its latest frontier AI model, designed with state-of-the-art cybersecurity and long-horizon capabilities. The model achieved the top rank on Artificial Analysis' updated Intelligence Index v5.0 with a score of 59. Gemini 4 Carbon demonstrates significant improvements in coding and agentic tasks, excelling in cybersecurity defense benchmarks. It scored 88.2% on SEC-bench Pro for vulnerability detection, 54.8% on ExploitGym for exploitation, and 71.0% on CWE-bench for patching. The model also shows enhanced code generation security and reduced misalignment and fabrication rates in live coding environments. Google is taking a phased approach to its rollout, initially making Carbon available to a subset of cyber defenders and trusted partners through its Fairwind Program, with broader access planned for developers, businesses, and consumers soon. Gemini 4 Carbon will be priced at $2 per million input tokens and $10 per million output tokens, the same introductory price as its predecessor, Gemini 4 Argon.

Gemini 4 Carbon's advancements in cybersecurity, particularly its high scores in vulnerability detection, exploitation, and patching benchmarks, signal a new era where AI models are capable of autonomously identifying and remediating security flaws. This suggests a market shift towards AI-powered security solutions that can operate at a scale and speed previously unattainable, potentially reducing the human effort required for threat hunting and patching.

The introduction of Gemini 4 Carbon, with its focus on long-horizon tasks and agentic workflows, indicates a move towards more sophisticated AI agents that can handle complex, multi-step engineering problems. The model's ability to persist through dead ends and its reduced misalignment and fabrication rates suggest that AI is becoming more reliable for critical tasks such as root-causing and patching subtle bugs across large codebases, impacting software development and maintenance practices.