CodeMender found the bodies
Google has launched Gemini 3.6 Flash alongside a new AI model built to find security flaws in code.
The model, called Gemini 3.5 Flash Cyber, is designed to be faster and cheaper than larger security models such as Anthropic’s Mythos 5.
It will first be available to governments and selected partners through CodeMender, Google’s AI security coding tool.
Because it costs less to run, CodeMender can use the model several times on the same task, helping it scan more code and uncover more vulnerabilities.
In brief:
Gemini 3.6 Flash: Better at coding and working with different types of content.
Gemini 3.5 Flash Cyber: Built for finding security weaknesses quickly.
Gemini 3.5 Flash-Lite: The cheapest model in the Gemini 3.5 range.
Hackers hate coupons
In testing, Flash Cyber found 55 confirmed issues in Google’s V8 JavaScript engine.
Gemini 3.5 Flash found 47, while Opus 4.6 found 36. Ten of Flash Cyber’s findings were missed by every other model tested.
Fifty-five security flaws later, someone at Google is quietly deleting “looks good to me” from Slack. - MG