Google launches a cheaper alternative to large AI security models like Mythos - The Verge

What the report says

The Verge reports that Google is introducing Gemini 3.5 Flash Cyber, a cybersecurity-focused AI model designed to find and help patch software vulnerabilities at lower cost than larger security models. The model is based on Gemini 3.5 Flash and is being positioned by Google as a less expensive alternative to systems such as Anthropic’s Mythos. The launch was described in a Google blog post cited by The Verge on Tuesday, July 21, 2026.

According to the report, Gemini 3.5 Flash Cyber will initially be made available to governments and trusted partners through CodeMender, Google’s security-oriented coding agent. Google says the model can be called repeatedly by CodeMender at high speed and low cost, which could let agents explore more code paths while searching for flaws.

The Verge places the announcement in a broader race among AI companies to develop models for cybersecurity work. Anthropic’s Mythos 5, released under its Project Glasswing effort, is described as a compute-intensive security model, and Microsoft has used Mythos in its own security checks. The report also notes that other companies, including Google and China’s Z.ai, are competing in this area.

Google also announced Gemini 3.6 Flash and 3.5 Flash-Lite. The company said 3.5 Flash Cyber showed competitive results on the CyberGym AI cybersecurity benchmark when invoked multiple times, and reported that it found 55 confirmed issues in the V8 JavaScript engine, including 10 not found by other tested models. Those figures are Google’s own claims as reported by The Verge.