Your first local chat with LM Studio

1. Install the right build

Use the Mac, Windows or Linux guide to check requirements and install LM Studio from its download page.

2. Find a small model

Open Discover. Search for a small chat or instruct model, such as Qwen2.5 0.5B, and choose a compatible GGUF variant. Check the actual file size against the memory guide, read the license and download it.

If that search has no suitable result, choose another small chat model offered by Discover. A search result does not guarantee that your machine can load it.

3. Load it in Chat

Open Chat, open the model loader at the top, and select your downloaded model. Let the app finish loading. Use a modest context size for the first run; increase it only after a short chat works.

4. Send a first message

Ask: “Explain a rainbow in three short sentences.” A successful load lets you send a message and receive a reply. After the model and runtime are downloaded, try another message with the internet disconnected.

5. Optional: import a Hugging Bay file

For a GGUF downloaded outside LM Studio, its docs describe the folder layout ~/.lmstudio/models/publisher/model/model-file.gguf. Place a compatible file in that layout under the app's configured models directory, then select it in Chat.

If the lms command is already available, the documented experimental shortcut is:

lms import "/path/to/model.gguf"

Replace the path with your downloaded file and follow the prompt. Importing does not convert an incompatible model. The import instructions explain the layout.