Skip to main content
1

Choose a device

First check Device Requirements to confirm the system, hardware, and available space.macOS and Windows download models inside the app; iOS and Android require preparing the model on a computer first.
2

Choose a model

For your first use, start with Edge0 8B: it is smaller and faster to download and load. When your device has enough resources, you can try Edge0 35B.Full model names:
  • Edge0-8B-A1B-preview
  • Edge0-35B-A3B-preview
For specifications, download links, and measured data, see Models: 8B vs 35B.
3

Prepare and load the model

Open Models, select a model, and click Download. When the download finishes, click Load model, then go to the chat page.
4

Send your first message

Go to the chat screen and enter:
Explain what a local language model is in three sentences.
Send it and wait for the reply. The first load and the first turn are usually slower than later ones. Once you see the model’s reply, you have completed your first local chat.Once a model is ready, responses are generated on your device without using a cloud model service. Downloading models and building the app still require a network connection; see Privacy & Data Handling for details.

​
No reply?

First confirm that the model is loaded, not just downloaded:
  • macOS: confirm you clicked Load model, and check on the Service page that the local service is running.
  • Windows: confirm model conversion has finished and that you clicked Enable.
  • iOS: confirm the selected model does not show Not included.
  • Android: confirm the model is imported, then tap Refresh and Load in order.
For further troubleshooting, see Troubleshooting.