Computer use becomes part of Gemini 3.5 Flash

Google DeepMind announced native computer-use support in Gemini 3.5 Flash. A previously separate capability becomes part of the main model, supporting agents that interpret interfaces and act in browser, mobile, and desktop environments.

Context

This broadens the applications an agent can operate without a dedicated integration. It also makes post-action checks more important. Seeing a button and clicking it does not establish that the operation finished or that the data was saved correctly. The system must observe the resulting state to close the loop.