HeadlinesBriefing favicon HeadlinesBriefing.com

Gemini 3.5 Flash Cyber: AI for Cybersecurity

Google DeepMind Blog •
×

Google DeepMind has introduced Gemini 3.5 Flash Cyber, a lightweight cybersecurity model designed to rapidly identify, validate, and patch software vulnerabilities. Built upon the efficient 3.5 Flash architecture, this specialized model offers a cost-effective alternative to larger, more expensive cybersecurity solutions.

Gemini 3.5 Flash Cyber is poised to enhance defender capabilities by scanning vast codebases and analyzing numerous codepaths more efficiently than mainline Flash models. This is particularly beneficial for complex tasks where extensive exploration is required. The model has demonstrated competitive performance against larger models on benchmarks like Cyber Gym and has surpassed mainline models in identifying critical vulnerabilities in evaluations by Google's Big Sleep team and within Google Chrome's production commit scanning pipeline.

Initially, Gemini 3.5 Flash Cyber will be available to governments and trusted partners through a limited-access pilot program via Code Mender, with broader expansion planned. This phased rollout aims to equip frontline defenders with advanced tools while mitigating potential misuse. Separately, foundational Code Mender capabilities will be integrated into generally available Gemini models through the Gemini Enterprise Agent Platform.