How to Install / Upgrade: Google's Gemini 3.6 Flash model cuts AI agent token costs by up to 65% on long horizon…

Google DeepMind released three new proprietary AI models it describes as among its most token-efficient yet: Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber. The models are aimed at making AI agents faster, smarter, and cheaper at scale. Gemini 3.6 Flash is reported to cut AI agent token costs by up to 65% on long horizon engineering tasks, and Gemini 3.5 Pro is on the way.

To install or upgrade, choose the model that matches your workload: Gemini 3.6 Flash for long horizon engineering agents where token cost is the main constraint, Gemini 3.5 Flash-Lite when you want a lighter Flash option, or Gemini 3.5 Flash Cyber when that Cyber variant fits the task. Update your agent or API client configuration so requests use the new model name instead of your current model, then redeploy or restart any services that hardcode the old model identifier. Review Google’s pricing for Gemini 3.6 Flash before rolling out at scale, and leave room in your config for Gemini 3.5 Pro when it becomes available.

Do not invent or assume model IDs, SDK versions, or release dates beyond what is published for these three models and the forthcoming 3.5 Pro. Confirm the live model names and pricing in your Google AI or Vertex AI console before production use. After switching, verify with a real long horizon engineering agent run that responses succeed under the new model name and that measured token usage and cost move in the expected direction relative to your previous setup, up to the stated 65% savings on that class of work for Gemini 3.6 Flash.