Google DeepMind has officially designated the Interactions API as the default interface for Gemini models and agents, replacing the legacy generateContent method in Google AI Studio. This generally available update, which followed a beta period starting in December 2025, means new agent features will now ship exclusively through this channel. While the older API remains functional for existing code, all future development must migrate to the new standard. A migration guide is available to assist developers with the transition.
The shift introduces significant structural changes to how applications interact with the models. Recent additions include Managed Agents running in their own Linux sandbox, background execution for long-running tasks, and tool chaining with Google Search and Maps. Google has also simplified the underlying schema by replacing the old role labels with typed steps that define every action from user input to function calls. Developers can now choose between Flex mode, which cuts costs by fifty percent, and Priority mode, which optimises for speed.
* Managed Agents operate within a dedicated Linux sandbox environment
* Flex mode reduces operational costs by half compared to standard settings
* Priority mode prioritises response speed over cost efficiency
