Pre-fetching grid blocks by user intent

Disclosed herein are systems, methods, and software to enhance user experiences with online services and hosted content. In at least one implementation, an initial user interaction with a user interface to a local model of content is detected. The local model of the content comprises an initial portion of the content downloaded to the local model. A subsequent user interaction with the user interface to the local model of the content is predicted based at least in part on the initial user interaction. A subsequent portion of the content is then identified for downloading to the local model of the content in anticipation of the subsequent user interaction occurring. The subsequent portion is identified based at least in part on the subsequent user interaction.