Models

LLMster lists models from your configured LM Studio server. Download new models on the host in LM Studio; the Apple app is where you choose which one to use for a chat.

Choose and load a model

1. Open the model picker from the model control in your conversation.
2. Search by name or use Sort, Format, and Capability to narrow the list.
3. Select a chat model. If a loading confirmation appears, choose Load Model.
4. Wait for loading to finish before sending your next message.

Understand model capabilities

Capability	What it enables
Vision	Image input, such as photos and screenshots.
Reasoning	Model reasoning that can be shown separately from the answer.
Tools	Requests to enabled integrations.

Capability information comes from the server. A model without vision cannot be made to understand images by attaching them, and a tool-capable model still needs configured integrations.

GGUF and MLX describe model formats. Availability depends on the host and LM Studio’s runtimes; choose from the models your own server supports.

Find a model again

Pin favorite models in the picker to keep them at the top. Model pins are remembered across launches and do not change the selected model by themselves. Refresh the picker after making changes on the host; use Try Again if retrieving the catalog fails.

Switch during a chat

Choose a different model from the same picker. Your visible chat remains, but model capabilities and available context can differ. Review any continuation warning before proceeding.

In Settings → App, Unload Previous Chat Model controls whether the previous chat model is unloaded before the new one loads. This setting preserves the configured summarizer model. To unload every model, use Utilities.

Models share the host's memory

Model size and loaded context both affect memory use. If loading fails, try a smaller model or free host memory. Unloading removes a model from memory, not from disk.

See model troubleshooting if a selected model is missing or never becomes ready.