Google is launching Gemini 3.5 Flash Cyber, a security-focused AI model intended to find and help patch software vulnerabilities at lower cost than larger systems. The Verge reports that the model will first be available to governments and trusted partners through CodeMender, Google’s security coding agent.
The model is built on Gemini 3.5 Flash and is designed to be called repeatedly at high speed and low cost. Google says that lets AI agents scan more code paths and search for more vulnerabilities than a slower, more expensive model might allow.
Google positions the system as an alternative to heavy security models such as Anthropic’s Mythos. In Google’s testing on the CyberGym benchmark, Gemini 3.5 Flash Cyber achieved competitive results when invoked up to five times. The company says it found 55 confirmed issues in the V8 JavaScript engine, including 10 that no other model found.
The announcement also arrived with Gemini 3.6 Flash and Gemini 3.5 Flash-Lite updates. For security teams, the key question will be how well repeated low-cost scans translate from benchmarks into real vulnerability triage.