Google Launches Gemini 3.6 Flash and Flash Cyber

Google released three new Gemini models on July 21, targeting faster AI agents, lower running costs, and software vulnerability detection.

Gemini 3.6 Flash and Gemini 3.5 Flash Lite are available now to developers and general users. Google is holding back the third model, Gemini 3.5 Flash Cyber, for governments and vetted partners because of its ability to locate and patch security flaws in code.

Flash 3.6 Cuts the Cost of Running AI Agents

Gemini 3.6 Flash handles coding, knowledge work, multimodal analysis, and multi-step AI agent tasks. Google states the model uses 17 percent fewer output tokens than its predecessor, with pricing set at $1.50 per million input tokens and $7.50 per million output tokens.

Coding performance also rose. Gemini 3.6 Flash scored 49 percent on the DeepSWE benchmark, up from 37 percent for Gemini 3.5 Flash, while making fewer unnecessary edits during test runs.

The release lands alongside continued build out of the infrastructure behind Google's AI products, including a multibillion dollar expansion of a US data center campus supporting Gemini workloads. That kind of physical build out carries its own strain, and recent reporting on the energy and water costs tied to AI data centers shows the tradeoff regulators are now watching closely.

Flash Lite Targets Bulk, Low-Complexity Work

Gemini 3.5 Flash Lite is built for tasks where speed and cost matter more than deep reasoning. Google says the model processes roughly 350 output tokens per second, priced at $0.30 per million input tokens and $2.50 per million output tokens.

Its intended uses include document processing, data extraction, agentic search, and automated subagent tasks. Flash Lite is live in the Gemini API, Google AI Studio, Android Studio, and the Gemini app, and Google has begun folding it into Search.

Flash Cyber Stays Behind a Locked Door

Gemini 3.5 Flash Cyber is a specialized model built to find, validate, and patch software vulnerabilities. It operates inside CodeMender, where multiple agents examine separate parts of a codebase before combining their findings into one security report. Google reports competitive results on the CyberGym benchmark.

Because vulnerability-detection tools carry dual-use risk, Google is limiting Flash Cyber to a pilot program for governments and trusted partners rather than a public launch. The same caution is shaping how companies deploy AI inside sensitive financial systems, including the debt now backing large-scale AI buildout described in recent reporting on bond sales tied to AI infrastructure.

What Comes Next

Gemini 3.6 Flash and Flash Lite are available immediately. Flash Cyber will roll out separately through its limited pilot. Gemini 3.5 Pro remains in partner testing with no confirmed public release date, and Google has started training Gemini 4 under what it calls its largest pre-training run to date.