Advanced AI Technologies

Google introduces Gemini 4 Carbon for cybersecurity

Google introduced Gemini 4 Carbon for cybersecurity and long-running tasks, reporting strong results on vulnerability-finding and patching benchmarks. The model was not yet publicly available.

Google introduced Gemini 4 Carbon, which it described as its strongest model for cybersecurity and long-running work. The model reportedly scored 88.2% on SEC-bench Pro, a benchmark for finding vulnerabilities, and 71.0% on CWE-bench, a benchmark for patching them.

The model was not yet available for general use at the time of the report. Its positioning indicates a focus on specialized software-security workflows rather than general conversational assistance.

Cybersecurity models can support vulnerability discovery and remediation, potentially helping security teams analyze code and address weaknesses more quickly. The reported benchmark results suggest an emphasis on both detection and repair capabilities.

The development matters because advanced AI systems are increasingly being evaluated on extended, task-oriented work and domain-specific performance. However, the source provides no independent validation of Google’s benchmark claims or details about pricing and deployment.

Original source: Newegg Insider

Related RAQI solutions

In-depth analysis on RAQI Insight

Work with RAQI