Open the menu and choose Models. Download a model once, then select it whenever you need it.
A simple starting point #
Try gemma3:4b for everyday text tasks. If it is too slow or too large for your computer, try the smaller gemma3:1b. Model speed and answer quality depend on your computer and the task.
The Ollama library currently lists downloads of about 3.3 GB and 815 MB respectively. Download size is not the same as memory needed while running. Browse Ollama’s model library for available versions and model licenses.
Download and switch #
- Enter a model name in Download a model.
- Choose Download model and wait for completion.
- Use Use model beside an installed model. You stay on the Models page and the saved selection is marked In use.
The selection applies to chat and to formulas that omit the optional model argument. Cached formula answers are reused when their inputs and model match.
Interrupted download #
If the registry connection closes or a download fails, choose Retry download. Ollama can reuse completed download parts. Existing downloaded models still work.
Local models only #
XLlama blocks cloud model names and verifies local model metadata before sending your content. Remote Ollama servers are not supported. Delete models through Ollama’s own tools if you need to free disk space.