Google's Gemini 3.5 Flash Cyber offers cheaper AI security than Mythos

Google has introduced Gemini 3.5 Flash Cyber, a dedicated AI security model designed as a cheaper alternative to heavyweight systems like Anthropic's Mythos. Built on the Gemini 3.5 Flash architecture, the model integrates directly with CodeMender, Google's security-focused coding agent. CodeMender can invoke 3.5 Flash Cyber multiple times at high speed and low cost, enabling it to scan more code paths and identify vulnerabilities that larger, more expensive models might miss. In benchmarks, 3.5 Flash Cyber achieved competitive performance against significantly larger models on the CyberGym AI cybersecurity benchmark when called upon up to five times. More notably, it identified 55 unique confirmed issues in the V8 JavaScript Engine, outperforming both Gemini 3.5 Flash (47 issues) and Opus 4.6 (36 issues). The model discovered 10 issues that no other model found, demonstrating its ability to uncover new code paths and vulnerabilities through repeated invocation.

Alongside the security model, Google also announced Gemini 3.6 Flash, which brings improvements in coding and multimodal performance, and Gemini 3.5 Flash-Lite, positioned as the most cost-effective model in the 3.5 series. The launch comes as the cybersecurity AI race heats up—Microsoft adopted Anthropic's Mythos for its security checks, leading to its biggest Patch Tuesday ever, while China's Z.ai claims its model can compete with Mythos. By offering a low-cost, high-speed alternative, Google aims to democratize AI-powered vulnerability detection, making it accessible to governments and trusted partners first through CodeMender. This could significantly accelerate the pace of security patching and reduce reliance on compute-heavy models.