Google has refreshed its Gemini model lineup with three new releases aimed squarely at production AI agents. Announced on the official Google blog as part of its July 2026 AI updates, the launch replaces Gemini 3.5 Flash with Gemini 3.6 Flash as the mid-tier workhorse, adds Gemini 3.5 Flash-Lite for bulk, low-cost tasks, and introduces Gemini 3.5 Flash Cyber, a narrow, security-focused variant.
Gemini 3.6 Flash is priced at $1.50 per million input tokens and $7.50 per million output tokens — about 17% cheaper than its predecessor — while posting double-digit gains on several coding, computer-use, and knowledge-work benchmarks. Independent coverage notes the model also brings faster responses and higher task-completion rates in early agentic testing.
The Flash tier matters because it carries the overwhelming majority of production API traffic: coding agents, chatbots, and automated workflows that call a model thousands of times a day, where per-call cost compounds quickly. Google is also rolling 3.6 Flash into GitHub Copilot, where it targets web and app development, longer-horizon agentic tasks, configurable reasoning effort, and parallel tool use.
For developers, the headline is token efficiency. A faster, cheaper Flash model with better benchmark scores lowers the cost of running agents at scale and makes Gemini a stronger default for high-volume pipelines. The dedicated Cyber variant signals where Google sees the next battleground: security-tuned models for alert analysis, incident summaries, and malware triage.