Introducing computer use in Gemini 3.5 Flash

What We Know

Google’s public materials and reporting say Gemini 3.5 Flash now includes “computer use” as a native capability. According to Google’s announcement and developer documentation, the feature lets the model inspect screenshots and take actions — for example deciding where to click or what to type — so agents built on Gemini can operate across browsers, desktop and mobile platforms. Coverage from multiple outlets frames this as the model being able to “see and operate your screen.”

Developer-oriented writeups and tutorials show how the Interactions API exposes these capabilities to builders: screenshots and action commands are part of the workflow, and journalists and blog posts describe the capability as enabling agentic automation (letting models complete multi-step tasks without continuous user intervention). Some reporting and community posts present early evaluations and tutorials demonstrating browser control, while other commentary describes the release as a preview rather than general availability in its current form.