Google Gemini 3.5 Flash Adds Built-In Computer Use

Google has integrated Computer Use directly into the Gemini 3.5 Flash model, allowing AI agents to control graphical interfaces without requiring a separate preview version. This change matters to developers and enterprise users who need reliable automation for tasks involving screenshots, mouse clicks, and keyboard input. The update removes the need to toggle between distinct models for general reasoning and desktop interaction.

Google consolidates desktop interaction into its fast coding model

The Gemini 3.5 Flash model serves as Google's latest fast model designed for coding and multi-step agent workflows. Google describes this version as its strongest coding model to date. The integration of Computer Use as a built-in tool streamlines the development process for agents that must navigate complex software environments.

The model supports a one-million-token input context window and can generate up to 65,000 output tokens. Computer Use enables agents to interact with graphical interfaces through screenshots, mouse clicks, scrolling, and keyboard input. Google has included adversarial training and specific safeguards to manage sensitive actions like financial transactions.

Users can access the feature through the Gemini API and Google’s Gemini Enterprise Agent Platform. The capability was previously available only as a separate preview model known as Gemini 2.5 or in earlier Gemini 3.x previews. This update consolidates those capabilities into the stable release.

When we covered the last gemini 3.5 flash update, several of the same balance and stability themes came up. The current release confirms that Google is moving Computer Use from a experimental preview into the core functionality of its fast model.

The feature is currently available globally through the specified platforms. This consolidation allows developers to build more robust agents that combine high-context reasoning with direct desktop control in a single stable model.