Where are Ollama models stored on a Mac?

In ~/.ollama/models , unless the OLLAMA_MODELS environment variable points somewhere else. Source: Ollama FAQ .

Where does LM Studio keep its models on a Mac?

LM Studio's documentation shows its models directory as ~/.lmstudio/models/ , one folder per publisher and model. You can change it in the My Models tab. Source: LM Studio docs .

Where is the Hugging Face cache on a Mac?

By default in ~/.cache/huggingface/hub . HF_HOME or HF_HUB_CACHE moves it. hf cache ls shows what is there. Source: Hugging Face Hub docs .

How do I see how big each model is?

ollama ls for Ollama, lms ls for LM Studio, hf cache ls for the Hugging Face cache. Each lists sizes.

Where local AI models live on a Mac: Ollama, LM Studio, Hugging Face

Ollama, LM Studio and Hugging Face each keep models in their own folder. Here is each folder, how to list what is in it with the tool's own command, and how to move or remove a model, from each tool's documentation.

Updated 3 Nov 2026. Every location and command here is from the tool's own documentation, linked where it is used, read on 3 Nov 2026. The one size was measured on the developer's Mac and carries its date.

Ollama, LM Studio and Hugging Face each keep downloaded models in a folder of their own. On the developer's Mac on 6 Oct 2026, ~/.ollama/models alone took 21.09 GB.

Ollama

Where: Ollama's FAQ, "Where are models stored?": "macOS: ~/.ollama/models". To use another folder, "set the environment variable OLLAMA_MODELS to the chosen directory." (Ollama FAQ)

List and remove (Ollama CLI reference):

ollama ls
ollama rm gemma4

Remove models with ollama rm, not by deleting files in blobs: models can share layers. More on Ollama's store.

LM Studio

Where: LM Studio's documentation shows "LM Studio's expected models directory structure" as ~/.lmstudio/models/, then a folder per publisher, then a folder per model, holding the model file (Import models). "LM Studio aims to preserves the directory structure of models downloaded from Hugging Face."

Move it: "You can change the models directory by heading to My Models" (Download an LLM).

List: "The lms ls command displays a list of all models downloaded to your machine, including their size, architecture, and parameters." (lms ls)

lms ls

Hugging Face

Where: "The default <CACHE_DIR> is ~/.cache/huggingface/hub." It can be changed "by specifying either HF_HOME or HF_HUB_CACHE environment variable." The same page describes an xet directory too, "located at ~/.cache/huggingface/xet by default". (Manage the cache)

List and remove, with the hf command from huggingface_hub:

hf cache ls
hf cache rm model/bert-base-cased
hf cache prune

Hugging Face: "Run hf cache ls to explore what is stored locally." "Use hf cache rm to permanently delete repositories or revisions from your cache." hf cache prune "removes shared files that no cached repo uses anymore."

All three at once

Each folder's total, from Terminal (a folder that does not exist prints an error and is skipped):

du -sh ~/.ollama/models ~/.lmstudio/models ~/.cache/huggingface

Before you remove a model

Getting it back means downloading it again. For a large model that takes time and bandwidth.
A model you made yourself (with ollama create, or an imported file) may not be on any server.
Removing a file frees nothing while it is still in the Trash, or while a Time Machine local snapshot holds it. Why.

Sources

Ollama FAQ and CLI reference, read 3 Nov 2026 (from github.com/ollama/ollama, docs).
LM Studio docs: Import models, Download an LLM, lms ls, read 3 Nov 2026 (from github.com/lmstudio-ai/docs).
Hugging Face Hub: Manage the cache, read 3 Nov 2026 (from github.com/huggingface/huggingface_hub, docs).

Where local AI models live on a Mac