Google releases Gemini 3.5 Flash Cyber security model

Google announced the release of Gemini 3.5 Flash Cyber, a new AI model built on Gemini 3.5 Flash and designed to locate and patch security vulnerabilities in code. The model is positioned as a lower-cost option compared with larger systems. A company blog post on Tuesday described it as a "cost-efficient and highly capable alternative" to more expensive AI security offerings.

Gemini 3.5 Flash Cyber will be available initially through CodeMender, the company's security-focused coding agent. The integration allows the model to be invoked multiple times at high speed and low cost, enabling broader scanning of code paths.

benchmarks On the CyberGym AI cybersecurity benchmark, the model achieved competitive results relative to significantly larger systems when called up to five times. In testing on the V8 JavaScript Engine, it identified 55 unique confirmed issues, compared with 47 found by Gemini 3.5 Flash and 36 by another model.

The model also discovered 10 issues that no other tested model found, according to the same announcement.

Gemini updates Google separately released Gemini 3.6 Flash, which includes improvements in coding and multimodal performance. A new 3.5 Flash-Lite variant was introduced as the most cost-effective option in the 3.5 series.