Google Rolls Out Gemini 3.6 Flash, 3.5 Flash-Lite and Flash Cyber

Google says it has launched three new “Flash” models for AI agent workflows: Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber. The upgrades target lower output cost, reduced token usage, and better built-in computer-operation capability to support large-scale enterprise deployments. Gemini 3.6 Flash is positioned as the main work model. Google highlights strong computer-use performance (OSWorld-Verified score 83.0%) and improved token efficiency, reporting 17% fewer output tokens vs. the previous version, and up to 65% fewer in tests such as DeepSWE. Pricing guidance cited: $1.50 per 1M input tokens and $7.50 per 1M output tokens. Gemini 3.5 Flash-Lite focuses on low latency and high throughput, with output speed up to 350 tokens/second and low input cost ($0.3 per 1M input tokens; $2.5 per 1M output tokens). It also supports adjustable “thinking levels,” and Google claims it can surpass the standard Gemini 3 Flash in long-context and agent coding evaluations. For security use cases, Google introduced Gemini 3.5 Flash Cyber, a model fine-tuned from 3.5 Flash. It is designed to collaborate with multi-agent tooling (CodeMender) to help identify and patch vulnerabilities. Due to dual-use concerns, access is limited to government entities and trusted partners via pilot programs. Google also notes Gemini 4 pre-training has begun, and developers can access the Gemini 3.6 Flash and 3.5 Flash-Lite via the Gemini API immediately.