Google adds native computer use to Gemini 3.5 Flash

Google has integrated a new "computer use" capability directly into its Gemini 3.5 Flash AI model, enabling the system to visually interpret and interact with computer screens across web, desktop, and mobile applications. The feature, which became available on June 24, 2026, allows developers to build intelligent agents that can perform actions such as clicking, typing, scrolling, and navigating software interfaces, advancing the automation of complex enterprise workflows.

Previously, this visual interaction capability existed as a separate, experimental model based on Gemini 2.5. Its integration into the core Gemini 3.5 Flash architecture streamlines development, allowing programmers to activate computer use as a built-in tool alongside other functions like code execution and search, without needing a separate AI model. Google says the enhancement moves the system beyond text-based reasoning and traditional function calling into full "agentic" computer interaction.

The technology operates through a screenshot-action loop. The model analyzes a screen capture, identifies interactive elements, and generates structured commands-such as mouse movements or keystrokes-to complete a...