Google shipped three new Gemini Flash models designed for developers and enterprise customers building AI agents: Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber.

Gemini 3.6 Flash cuts token usage by 17% compared to its predecessor while reducing costs per output token. The model shows improved precision in coding, knowledge work, and multimodal tasks including document parsing and data analysis.

Gemini 3.5 Flash-Lite delivers the fastest output in the series at 350 tokens per second, targeting high-volume, cost-conscious workflows. It outperforms earlier Flash-Lite models in coding and agentic tasks.

The cybersecurity-focused Gemini 3.5 Flash Cyber is fine-tuned for vulnerability detection and patching. Google restricts its deployment to governments and trusted partners through CodeMender in a limited pilot to prevent misuse.

Enhanced safety measures

Google strengthened safety safeguards across the models, particularly in domains like CBRN and cyber offense. The company said this reduces risks of misuse as AI models increasingly find vulnerabilities faster than teams can fix them.

Gemini 3.6 Flash and 3.5 Flash-Lite are available immediately through the Gemini API and Gemini Enterprise. The cybersecurity variant remains in restricted access.

Google continues iterating on the Gemini family, with 3.5 Pro in partner testing and Gemini 4 in pre-training. The releases reflect the company's focus on AI agent infrastructure and production reliability.