Particle.news

Google Expands Gemini With Flash Line as Pro Variant Stays in Testing

Reduced token use, lower latency and a security-focused build aim to cut developer costs and speed agent workflows.

Overview

  • Google announced Tuesday three new Gemini models — Gemini 3.6 Flash, 3.5 Flash-Lite and 3.5 Flash Cyber — targeted at token efficiency, low-latency agent use, and code-security review respectively.
  • Gemini 3.6 Flash is claimed to produce better inferences while using roughly one-sixth fewer tokens than Gemini 3.5 Flash, which Google says will lower token costs for users.
  • Gemini 3.5 Flash-Lite is tuned for lower latency to power AI agents and faster interactions, trading some generality for quicker responses in agent workflows.
  • Gemini 3.5 Flash Cyber is a restricted, security-focused variant built to work with Google’s Codemender agent and will be available only to government agencies and selected institutional partners.
  • Google continues partner testing of the higher-accuracy Gemini 3.5 Pro with no public release date, and independent analysis warns that growing AI deployments could substantially raise U.S. data-center electricity use through 2035, creating infrastructure and policy challenges.