Where apps hide your files

LM Studio's models directory

Updated 2026-10-28 · 5 min read

LM Studio downloads models into a hidden .lmstudio folder in your home directory, organised by publisher and model name. You can point it at an external drive from inside the app.

Where the models are

~/.lmstudio/models/

That is the models directory LM Studio's documentation gives, and the layout under it follows publisher/model/file, for example:

~/.lmstudio/models/infra-ai/ocelot-v1/ocelot-v1-instruct-q4_0.gguf

(That example path comes from LM Studio's own import documentation.) The folder is hidden because of the leading dot. In Finder, press ⇧⌘G, paste ~/.lmstudio/models and press Return, or from Terminal:

open ~/.lmstudio/models

If ~/.lmstudio is missing or empty and you have used LM Studio for a long time, check whether the models directory was changed (next sections) and whether an older install left files under ~/.cache/lm-studio, a location several third-party guides give for early versions:

ls ~/.cache/lm-studio

How the folder is laid out

LM Studio keeps the folder structure that Hugging Face uses, so the publisher folder is the Hugging Face user or organisation, the next folder is the model repository, and the files inside are the model weights. For most models that is a .gguf file per quantization: a label such as q4_0 or Q4_K_M is the quantization level, and higher numbers mean bigger, higher quality files. Several quantizations of one model sit side by side in the same folder. Models in Apple's MLX format, which LM Studio runs on Apple silicon Macs, arrive as a folder of weight and config files rather than one file.

Next to models, the ~/.lmstudio folder holds other things LM Studio keeps. The documentation puts saved chats at:

~/.lmstudio/conversations/

Chats are JSON files, and LM Studio advises against editing them by hand or relying on their structure. In the app you can right-click a chat and choose Reveal in Finder to jump to its file. The other subfolders vary between versions, so list them rather than relying on a name:

ls -a ~/.lmstudio

Change the models directory

LM Studio's documentation says you change the models directory by heading to My Models. Open it, find the models directory control, and choose a new folder, for example on an external SSD. A few practical points:

1. Unload (eject) any model that is loaded.
2. Move your existing models into the new folder before you switch, keeping the publisher/model/file structure. A file dropped in without the two folder levels may not appear in the list.
3. After switching, check My Models to see that every model is listed.
4. Make sure the drive is mounted when you launch LM Studio. With it unplugged the list will look empty, but the models are not gone.

macOS may ask permission the first time an app reads an external volume. Allow it, or the folder will appear empty.

Adding models you already have

If you downloaded a .gguf file elsewhere, LM Studio documents a command line route, marked experimental:

lms import <path/to/model.gguf>

It walks you through a short prompt and places the file in the right publisher/model structure, which is safer than dragging files into the folder yourself.

Space and cleaning up

A single model file can be anything from under a gigabyte to tens of gigabytes, and each quantization you download is another file. To see the total:

du -sh ~/.lmstudio/models

To free space, remove models from My Models inside LM Studio, so the app's list stays in step with the disk. If you do it in Finder instead, quit LM Studio first, move the model's file (or its MLX folder) to the Trash, and remove the now empty model folder as well. Moving a model to the Trash is reversible until you empty it. Do not delete the whole ~/.lmstudio folder unless you also want to lose your chats and settings.

Sharing models with other tools

LM Studio's models and Ollama's models are stored differently. Ollama uses hashed blob files with manifests, so you cannot point one tool at the other's folder. Tools that read plain .gguf files can often use a copy of an LM Studio file, but check that tool's own documentation. Hugging Face's Python libraries keep a separate cache with its own layout, so a model downloaded through Python is not automatically visible in LM Studio, nor the other way round.

Backing up or moving to a new Mac

Models can be downloaded again, so the cheapest backup is a list of the ones you use. If you would rather copy them, copy ~/.lmstudio/models and keep the folder structure, then open My Models on the new Mac. Chats in ~/.lmstudio/conversations are worth copying separately because they cannot be downloaded again.

Why Spotlight misses all of it

~/.lmstudio is a hidden folder, and Spotlight does not show results from hidden folders. A Spotlight search for a model name finds, at best, the model's web page.

Finding a model file by name

Quantized files carry the model and quantization in their names, so those are the terms to type. Typing q4_k_m, or .gguf together with part of the model name, into Everywhere lists matching files in the hidden .lmstudio folder; copies in Downloads or on an external models drive, even one you have since unplugged, join them once it has Full Disk Access. Press ⌘↵ to reveal one in Finder. It matches names only and does not look inside a model.

Common questions

Where does LM Studio store models on a Mac?

In ~/.lmstudio/models/, organised as publisher, then model, then the model file (usually .gguf). The folder is hidden in Finder, so open it with Go to Folder (⇧⌘G).

How do I change the LM Studio models directory?

LM Studio's documentation says to change it from My Models. Move your existing models into the new folder first, keeping the publisher/model structure, and keep that drive mounted when you launch the app.

Where does LM Studio save chats on macOS?

In ~/.lmstudio/conversations/ as JSON files. Right-click a chat and choose Reveal in Finder to find its file.

Can I import a model I downloaded separately into LM Studio?

Yes. The documented command is lms import <path/to/model.gguf> (marked experimental), which places the file in the right folder structure.