The LM Studio models folder on Mac: find, move and delete models

The LM Studio models folder on a Mac is ~/.lmstudio/models, a hidden folder in your home folder, with a folder for each publisher and, inside it, one for each model.

Earlier versions of LM Studio used ~/.cache/lm-studio/models instead, and you can point LM Studio at any other folder. The quickest way to be sure is to ask LM Studio itself.

Open My Models in LM Studio. Next to Models Directory, the path in use is shown with a menu button: Reveal in Finder opens the folder, and Change… points LM Studio at another folder, for example on an external drive.

To delete a model, use the same tab: click the menu next to the model and choose Delete. LM Studio downloads it again from Discover if you want it back. Models you imported from elsewhere are different, because the copy in LM Studio's folder may be the only one you have.

This post covers what's inside the folder, how big it gets, how to move it to an external drive without breaking anything, and what else LM Studio keeps next to your models.

Where the LM Studio models folder is on a Mac

LM Studio keeps everything in one folder in your home folder, named ~/.lmstudio. The dot at the start makes it hidden, so you won't see it in Finder until you go there directly.

That's also why it's easy to overlook. Nothing in your Applications folder or Documents hints that tens of gigabytes of models sit in your home folder, and dragging LM Studio to the Trash leaves them there.

What's the default location?

New installs keep models in ~/.lmstudio/models. LM Studio's documentation describes the layout: the models folder holds one folder per publisher, each publisher folder holds one per model, and the model's files sit inside that.

The layout copies Hugging Face, where most models come from. A model published as infra-ai/ocelot-v1 on Hugging Face ends up in a folder named infra-ai, then ocelot-v1, with its file inside.

Why might your models be somewhere else?

There are two common reasons. If you've used LM Studio since its earlier versions, your models may still be in ~/.cache/lm-studio/models, the folder those versions used. And if you, or a guide you followed, changed the models directory, they're wherever you pointed it.

Some Macs end up with both folders, each holding models. LM Studio only lists the models in the folder it's set to, so the others are easy to forget.

How do you open the models folder in Finder?

The reliable way is through LM Studio, because it shows the folder actually in use:

1. Open LM Studio and click My Models in the sidebar, or press Command-3.
2. Find Models Directory near the top, with the folder's path next to it.
3. Click the menu button beside the path and choose Reveal in Finder.

You can also go there without LM Studio. In Finder, choose Go › Go to Folder, type ~/.lmstudio/models and press Return. For an older install, try ~/.cache/lm-studio/models as well.

What else is in the .lmstudio folder?

Models are the largest part by far, but not the only one. LM Studio's documentation names several other folders, and some hold things you'd want to keep.

Folder	What it holds	Safe to remove?
~/.lmstudio/models	Downloaded and imported models	Model by model, in LM Studio
~/.lmstudio/conversations	Your chats, as JSON files	Only if you don't want the chats
~/.lmstudio/config-presets	Presets you saved	They're small; keep them
~/.lmstudio/hub	Presets shared on or downloaded from the LM Studio Hub	They're small; keep them
~/.lmstudio/bin	LM Studio's command-line tool	Only with the app
~/.cache/lm-studio/models	Models from earlier versions	Check what's there first

What's inside an LM Studio model folder

Open a publisher's folder and you'll see one folder per model, often with names that say how the model was packaged. The packaging decides how many files there are and how large they get.

What do the publisher names mean?

The first level is whoever uploaded the files to Hugging Face, which isn't always who made the model. Names such as lmstudio-community, mlx-community or bartowski are people and groups who convert models into formats LM Studio runs.

That's why one model can appear under several publishers. A model you downloaded twice from different uploaders is two full copies, even if it's the same model underneath.

What's the difference between GGUF and MLX models?

LM Studio runs two kinds of model on a Mac. GGUF models run on llama.cpp and are usually one large file ending in .gguf. MLX models run on Apple's MLX framework, which LM Studio supports on Apple silicon, and are a folder of weight files with small configuration files beside them.

Both work, and neither is smaller as a rule. What matters for space is that the same model in both formats is stored twice. If you tried both and settled on one, the other is a candidate to delete.

Why are there several versions of the same model?

When you download a model, LM Studio offers several options with names such as Q3_K_S or Q_8. LM Studio's documentation explains that these are copies of the same model at different levels of quantization, a way of compressing the file at some cost to quality.

The documentation suggests choosing a 4-bit option or higher if your Mac can run it. Each option you download is a separate file of its own size, so trying three quantizations of one model keeps three files.

Which files in a model's folder matter?

For a GGUF model, the .gguf file is the model. A model's folder can hold several of them side by side, one per quantization you downloaded, and each can be removed on its own. Very large models are sometimes split into numbered parts, which only work together.

Vision models often come with a second, smaller file that handles images. Keep it in the same folder as the model it belongs to. For an MLX model, the whole folder is the model; don't remove single files from it.

What about models you imported?

You can use models downloaded outside LM Studio by placing them in the same publisher and model structure inside the models folder. LM Studio's command-line tool can also import a model file for you.

Read how that tool works before you delete anything. LM Studio's documentation says it moves the file into the models folder by default, unless you ask it to copy, hard-link or symbolically link it instead. A model imported the default way may now exist only in LM Studio's folder.

How much space LM Studio models take

A single model is usually several gigabytes, and a handful of experiments adds up quickly. Because nothing is removed automatically, the folder holds every model you've tried, including ones you used once.

How big is one model?

It depends on the number of parameters and the quantization. LM Studio's own documentation lists sizes for a few popular models, and they're a fair picture of the small ones.

• Qwen2.5 7B, MLX 4-bit4.3 GB · Review first
• Ministral 8B, GGUF4.7 GB · Review first
• Qwen2.5 7B, GGUF4.7 GB · Review first
• Llama 3.1 8B, GGUF4.9 GB · Review first
• Llama 3.1 8B, MLX 8-bit8.5 GB · Review first

Download sizes LM Studio's tool-use documentation lists for each model. Larger models, with more parameters or less compression, run to tens of gigabytes each.

Llama 3.1 8B above is nearly twice as large in its 8-bit MLX version as in its GGUF download. Models with tens of billions of parameters take tens of gigabytes even at 4 bits.

How do you check a model's size before downloading?

In Discover, select a model and look at its download options. In the versions LM Studio's documentation shows, each option lists its size, and the Download button repeats it. The same screen says whether the model fits in your Mac's graphics memory.

That's the moment to decide. A model of 5 GB needs 5 GB free to download, and it stays until you delete it.

How do you see the total?

LM Studio tells you. In the versions its documentation shows, My Models opens with a line saying how many local models you have and how much disk space they take, and LM Studio's command-line tool prints the same summary. The example in the documentation reads 47 models taking 160.78 GB.

In Finder, select the models folder and choose File › Get Info for its total. To see which models are largest, open the folder, choose View › as List, then View › Show View Options and select Calculate all sizes.

Why don't Finder and LM Studio agree?

Usually because of imported models. A hard-linked model is counted in full in both places it appears, though the data is stored once. A symbolically linked one shows up in LM Studio while its file lives elsewhere, so the models folder can look smaller than LM Studio's total.

The other reason is a second models folder. If LM Studio is set to one folder and an older one still holds models, Finder counts both while LM Studio lists only one.

Do you have the same model twice in other apps?

Quite possibly. LM Studio keeps its own copies, separate from Ollama's models and from Hugging Face's cache, even when the model is the same. Someone who tried a model in each tool has downloaded it three times.

Ollama stores models in its own layered format; the post on deleting Ollama models explains it. The guide to local AI models covers where the other tools keep theirs.

Does running a model use more disk space?

Not much. LM Studio's documentation explains that loading a model means placing its weights in your Mac's memory, not making a second copy on disk. The disk cost is the file itself.

There's one indirect cost. If a model needs more memory than your Mac has free, macOS moves other data out of memory into swap files on the startup disk. Those go away after a restart, but a nearly full disk leaves less room for them.

Does each account on the Mac have its own models?

Yes. The models folder is inside each person's home folder, so two accounts that download the same model store it twice. If several people use LM Studio on one Mac, check each account's folder when you're looking for space.

How to delete models in LM Studio

Deleting through LM Studio is the cleanest route for a single model: it removes the files and updates its list in one step. For a large cleanup, Finder works too, as long as LM Studio is closed.

Delete a model in My Models

1. If the model is loaded in a chat or on the local server, unload it first.
2. Click My Models in the sidebar.
3. Find the model, then click the menu button next to it.
4. Choose Delete and confirm.

The labels have moved a little between versions, but the menu next to each model is where the option lives. LM Studio's command-line tool lists, downloads, loads and imports models, but its documented commands don't include deleting one.

Can you delete models in Finder instead?

Yes. The models folder is a plain folder of files, and moving a model's folder to the Trash removes that model and nothing else. It's the faster way to clear several models at once.

1. Quit LM Studio, so it isn't using any of the files.
2. Open the models folder from My Models with Reveal in Finder, or with Go › Go to Folder.
3. Open a publisher's folder and move the model folders you no longer need to the Trash.
4. If a publisher's folder is now empty, move it to the Trash too.
5. Open LM Studio and check My Models, then empty the Trash when you're sure.

Until you empty the Trash, the space isn't free, but you can put a model back with File › Put Back if you change your mind.

Why are empty folders left behind?

A report on LM Studio's bug tracker describes deleting a model in My Models and finding its folder still on disk, empty. An empty folder takes next to no space, so it's harmless. You can move it to the Trash in Finder if you like a tidy folder.

Can you delete the whole models folder?

You can, with LM Studio closed, and it removes every model at once. Don't remove the ~/.lmstudio folder itself just to clear models: it also holds your chats and presets, which aren't stored anywhere else.

If you only want to start over, move the publisher folders to the Trash and leave the models folder in place. LM Studio fills it again as you download.

What deleting a model costs

Deleting a model never touches your chats or presets. What it costs depends on where the model came from.

What you delete	What you lose	How you get it back
A model downloaded in LM Studio	A download, often several GB	Discover, the same model and quantization
A model you imported by moving it	Possibly the only copy	From wherever it came from, if it still exists
A model imported as a copy	The copy only; the original stays	Import it again
A hard-linked model	No space until the other link goes too	Import it again
An unused quantization or format	Nothing, if you use another	Discover

Why didn't deleting a model free any space?

Three reasons cover almost every case. A model you removed in Finder sits in the Trash until you empty it. A hard-linked model frees nothing until its other link is gone too. And if you use Time Machine, local snapshots can hold deleted files for a while.

Snapshots clear themselves as they age or when space is needed. The guide to Time Machine's local snapshots explains how to check for them.

How long does a model take to come back?

As long as it takes to download several gigabytes from Hugging Face. LM Studio needs an internet connection to search for and download models, though it runs downloaded models offline. On a slow or metered connection, keep the model you'll want next week.

What about models that are no longer available?

Uploads on Hugging Face can be renamed, replaced or removed. If a model matters to you, such as a fine-tune you made or one a colleague shared, keep a copy of its file somewhere other than LM Studio's folder before you delete it.

Do chats that used the model still work?

Your chats stay in their own folder whatever you delete. To carry on a conversation, load another model; the earlier messages are still there for it to read.

Moving the models folder to an external drive

If you want to keep your models but not on the startup disk, LM Studio can use a folder anywhere, including an external drive. The setting is in My Models, next to Models Directory.

Move the models folder, step by step

1. Unload any model you have loaded, then quit LM Studio.
2. On the external drive, create a folder for your models.
3. In Finder, copy the publisher folders from your current models folder into the new one.
4. Open LM Studio, click My Models, click the menu button next to Models Directory, and choose Change….
5. Select the new folder. Check that your models appear and that one loads.
6. Move the old publisher folders to the Trash, then empty it when you're sure.

Copying first and deleting last means nothing is lost if the drive or the copy has a problem. LM Studio's documentation notes that its command-line tool uses whichever models directory you set in My Models.

Can the models folder live in iCloud Drive or Dropbox?

It can, but it's a poor fit. A cloud folder uploads every model you download, which uses your cloud storage and your connection, and it may keep only a placeholder on the Mac when space is short. LM Studio then has to wait for a download before it can load the model.

Your default models folder isn't synced, because ~/.lmstudio sits outside the Desktop and Documents folders that iCloud can store. Keep it that way, and pick a local or external folder when you use Change….

Which drive works best?

A fast drive that stays connected. LM Studio reads a model from disk each time it loads it, so a slow drive means slow loads. A drive formatted as APFS suits a Mac-only setup; Disk Utility can format a new drive before you start.

When the drive isn't connected, LM Studio can't see the models on it. Downloads go to whichever folder is set, so new models will fill the external drive, not your startup disk.

How do you move back to the default folder?

Copy the publisher folders back to ~/.lmstudio/models, then choose Reset to Default Path from the same menu in My Models. Check your models are listed before you remove the copies on the drive.

What about moving to a new Mac?

Models are just files, so they move like any others. Migration Assistant brings your home folder across, hidden folders included. To copy by hand instead, open LM Studio once on the new Mac and quit it, then put the publisher folders into its models folder. LM Studio lists them the next time it opens.

If your models live on an external drive, connect it to the new Mac and choose the folder with Change… in My Models.

Should you use a symbolic link instead?

There's no need. Guides written for other tools suggest replacing the folder with a link to the drive, but LM Studio has its own setting for the folder's location. A link breaks quietly when a drive is renamed or reformatted, while the setting is the documented way and LM Studio shows it in My Models.

Backups and the rest of LM Studio's folder

Models are large, change rarely and can usually be downloaded again, which makes them a poor fit for a backup that runs every hour. The smaller folders beside them are the opposite.

Should you exclude models from Time Machine?

It's worth considering. Every model you download is copied to your backup disk, and models you delete stay in older backups for a while. Excluding the models folder keeps backups smaller.

1. Choose Apple menu › System Settings, click General, then Time Machine.
2. Click Options, then click the add button and choose the folders to exclude. You can also drag folders into the list.
3. Click Done.

The steps are the same in macOS 27 and macOS 26. Apple notes that excluded items still appear in Time Machine's local snapshots. Keep imported models you can't download again backed up somewhere else first.

Exclude the models folder, not the whole ~/.lmstudio folder. Your chats and presets are small and can't be downloaded again, so they're exactly what a backup is for.

Where are LM Studio's chats stored?

In ~/.lmstudio/conversations, as JSON files. LM Studio's documentation suggests Control-clicking a chat and choosing Reveal in Finder to find it, and advises against editing the files by hand. Chats aren't a cache: delete them in LM Studio if you want them gone.

What are presets and the hub folder?

Presets are saved system prompts and settings, kept in ~/.lmstudio/config-presets. Presets you share on the LM Studio Hub, or download from it, go to ~/.lmstudio/hub. Both are small text files and not worth removing for space.

What about LM Studio's runtimes?

LM Studio runs models with engines it calls LM Runtimes, such as llama.cpp and MLX, and downloads updates to them separately from the app. To see them, press Command-Shift-R in LM Studio. LM Studio needs the runtime for each model format you use, so manage them there rather than in Finder.

Uninstalling LM Studio completely

Dragging LM Studio to the Trash removes the app and nothing else. Your models, chats and presets stay in ~/.lmstudio, and in any other folder you set as the models directory, until you remove them.

What should you remove, and in what order?

1. Open My Models and note the Models Directory path, in case it's on another disk.
2. Save any chats you want to keep, from ~/.lmstudio/conversations.
3. Quit LM Studio.
4. Move LM Studio from your Applications folder to the Trash.
5. Move ~/.lmstudio to the Trash, plus the models folder you noted, if it was elsewhere.
6. Check for ~/.cache/lm-studio from an older version, and remove it too.

The guide to uninstalling apps properly covers the smaller support files apps leave in your Library folder.

Does LM Studio's command-line tool need removing?

No separate step is needed. LM Studio's documentation shows the tool inside ~/.lmstudio/bin, so it goes when that folder goes.

Should you keep the models when you uninstall?

If you might use them in another tool, yes. A GGUF file also runs in other apps built on llama.cpp, so move the model files you want to keep to another folder before you remove ~/.lmstudio. Everything else in that folder is specific to LM Studio.

Choosing which models to keep

Most people keep a few models they use and a long tail they tried once. The tail is where the space is.

A quick way to decide

The model	Keep it?
One you use every week	Yes
A second quantization of a model you kept	Delete the one you don't use
The same model in both GGUF and MLX	Keep the format that runs better for you
A model you tried once and didn't like	Delete it
One you imported or made yourself	Keep a copy elsewhere first
An older version of a model with a newer release	Delete it once the new one works

What about embedding models?

LM Studio lists embedding models, which turn text into numbers for searching documents, apart from language models. They're small: the examples in LM Studio's documentation are 24.81 MB and 84.11 MB. They're rarely worth deleting for space, and something you set up may rely on them.

Can your Mac run the models you're keeping?

A model too large to load is space with no use. LM Studio runs only on Mac computers with Apple silicon and recommends 16 GB of memory or more; on 8 GB Macs, its documentation suggests sticking to smaller models and modest context sizes.

As a rough guide, a model needs at least as much free memory as its file size, plus room for the conversation. A 40 GB model you downloaded on a whim may never load on a 16 GB Mac.

A worked example

Say My Models lists Qwen2.5 7B in both GGUF and MLX 4-bit, and Llama 3.1 8B in both GGUF and MLX 8-bit. Using the sizes in LM Studio's documentation, that's about 22 GB for two models.

Keep one format of each. Dropping the Qwen2.5 GGUF and the Llama 8-bit MLX version keeps both models and frees about 13 GB. Delete any model you tried once and didn't take to, and the folder shrinks again.

Mistakes to avoid

• Deleting the .lmstudio folder to free space. It takes your chats and presets with the models.
• Assuming an imported model can be downloaded again. LM Studio's import moves files by default.
• Changing the models directory without copying first. LM Studio then lists no models, and it looks as if they've gone.
• Clearing models while one is loaded. Unload it first, or quit LM Studio before using Finder.

Keeping it from growing back

Delete a model as soon as you've decided against it, rather than keeping it for later. Try a model at one quantization before downloading others. Once a month, open My Models, read the total at the top, and remove what you haven't loaded since last time.

In Finder, list view with the Date Modified column shows roughly when each model arrived. Models downloaded a year ago and never replaced are often the ones you've stopped using.

How the app does it

Wombit explains LM Studio's models folder like everything else on the disk. Select it in the map, the app's home screen, and the inspector says what it is, that LM Studio made it, and how its models are removed, from the catalog's entry for LM Studio models. It reads only names, sizes and dates, and file names and paths never leave your Mac.

• ~/.lmstudio/models

  LM Studio models

  Language models downloaded or imported in LM Studio, in a folder for each publisher and model.

  Review first · Moves it to the Trash

The folder is Review first: models are your own choice to keep, so it never starts checked and never counts toward the safe-to-remove total. Both the current folder and the one earlier versions used are recognized. A models folder you moved elsewhere with Change… isn't at either path, so the catalog's entry doesn't cover it.

Checking it in Cleanup moves the whole folder, every model in it, to the Trash after the review sheet, where you can put it back. To remove a single model, use LM Studio's My Models, as above.

LM Studio's chats and presets aren't in the catalog. For folders like those, the inspector can give a Best guess, written on your Mac by Apple Intelligence and labelled as a guess.

Seeing what LM Studio's models take is free. Moving them to the Trash is PRO, one $4.99 payment with lifetime access. Every Pro tool's review is free to open, so you see what would go and why before you buy.