Google ships Gemini 3.6 Flash, 3.5 Flash-Lite, and cybersecurity-focused Flash Cyber

Google DeepMind shipped a slate of new Gemini models: Gemini 3.6 Flash (more capable and token-efficient), Gemini 3.5 Flash-Lite (low-cost, high-volume inference), and Gemini 3.5 Flash Cyber, a cybersecurity-specialized model built on 3.5 Flash. Flash Cyber is designed to discover, validate, and patch software vulnerabilities, available to governments and trusted partners through CodeMender in a limited-access pilot.

DeepMind claims 3.5 Flash Cyber outperforms Gemini 3.5 Flash, 3.6 Flash, and Anthropic's Claude Opus 4.6 at identifying new vulnerabilities, and reports frame it as rivaling Anthropic's Mythos security model. The lineup targets AI agents and cost-sensitive scaled deployment — reflected in Alphabet Q2 earnings, where Sundar Pichai noted model APIs process 22 billion tokens/minute (up from 16B), driven by workhorse Flash models, with the Gemini app reaching 950M monthly active users.

The conspicuous absence: Google's flagship Gemini Pro model was not part of this release, a gap that fueled 'where's Pro' speculation, especially against Opus 5 and GPT-5.6 launching the same week. Separately, Bloomberg reported Google is developing an internal server chip, 'Frozen v2,' optimized for Gemini and targeted for 2028.

DeepMind CEO Demis Hassabis also floated a 'FINRA for AI' self-regulation proposal, which gained momentum but divided commentators over whether industry-led oversight is genuine accountability or regulatory capture. Watch for the Pro model's eventual release and whether Flash Cyber's vulnerability-patching claims hold up in the CodeMender pilot.