Gemini 3.5 Flash gains integrated computer capabilities

Google integrates computer use into Gemini 3.5 Flash, enabling developers to build agents that observe, reason, and act across web and desktop interfaces.

Google is integrating computer use directly into Gemini 3.5 Flash, whereas this capability was previously offered via a separate model, Gemini 2.5 computer use.

The idea: to enable developers to create agents capable of observing an interface, reasoning about what they see, and acting within web, mobile, or desktop environments. All of this adds to Gemini's existing capabilities, such as function calling, Search, or Maps grounding. Google highlights extensive and professional use cases, including continuous software testing, business application automation, and certain knowledge work tasks. Regarding security, 3.5 Flash benefits from targeted training against prompt injections, with two optional safeguards for enterprises: requiring explicit confirmation before a sensitive action and automatically interrupting a task in case of a suspicious indirect instruction.

Access is available via Gemini API, Gemini Enterprise Agent Platform, a Browserbase demo, and a reference implementation published by Google.