Answers · System 2

The model Puckie uses to answer questions and chat. It runs on your Mac by default. A bigger one gives better answers.

The built-in model

Qwen3-4B, running privately on this Mac. Puck starts it for you. Nothing to set up.

A model on another Mac

Have a second Mac with more memory? Run a bigger model there and Puckie will use it over your home network.

On the other Mac

1. Install Ollama and open it.
2. Download a model. In Terminal: ollama pull qwen3:8b (or any model you like).
3. Click the Ollama icon in the menu bar → Settings → turn on Expose Ollama to the network.
4. Find the Mac's name: System Settings → General → Sharing → Local hostname, for example studio.local.

In Puck

1. Menu bar icon → Settings… → System 2 → Another model.
2. Click Another Mac, then replace YOUR-MAC with that Mac's name.
3. Click Find models, choose one, then Try it. When it answers, click Save.
4. The first time, macOS asks whether Puck may find devices on your network. Click Allow.

LM Studio instead of Ollama

In LM Studio, load a model, open the Developer tab, start the server and turn on Serve on local network. In Puck, use LM Studio (on this Mac) or Another Mac and change the port to 1234.

An online model

Any service that works with OpenAI's API can answer for Puckie: enter its address, model and key. Puck warns you when what you say would leave your network.

If it doesn't work

Puck says	Try
No Mac called … on this network	Check the name in Sharing on the other Mac, or use its IP address.
No answer from …	Is the other Mac awake? Is Expose Ollama to the network on?
Found …, but no model is being served	Open Ollama (or LM Studio's server) on that Mac.
macOS is keeping Puck off the local network	System Settings → Privacy & Security → Local Network → turn on Puck.