ollama, mac, disk space

How do I delete Ollama models on a Mac and free disk space?

Last updated 2026-11-12 · 3 min read · By Muhammad Hamza

Run ollama ls to see your models, then ollama rm to remove the ones you no longer use. Where Ollama stores models on macOS, how to see what is big, and how to remove them safely.

To delete an Ollama model on a Mac, run ollama ls to see what you have, then ollama rm <model> for each one you no longer use. Ollama keeps models in ~/.ollama/models on macOS, and that folder is usually where the space went.

This guide shows how to find that folder, see what is big, remove models with the official command, and avoid the one mistake that costs people an evening: deleting a model that takes an hour to download again.

Where does Ollama store models on a Mac?

On macOS, Ollama stores models in ~/.ollama/models. That is /Users/<you>/.ollama/models, and the leading dot makes it a hidden folder in Finder. Press Cmd+Shift+. in Finder to show hidden folders. This path comes straight from the Ollama FAQ.

If someone set the OLLAMA_MODELS environment variable, models live in that folder instead. The FAQ says that variable is how you point Ollama at a different directory.

How do I see how much space Ollama is using?

Open Terminal and run:

du -sh ~/.ollama/models
ollama ls

The first command prints the total size of the folder. The second lists every model Ollama knows about. If du prints a number that surprises you, the list tells you which models to look at first. Models with a big parameter count are the usual culprits.

How do I delete a model?

Use the command from the Ollama CLI reference:

ollama rm llama3.2

Replace llama3.2 with the name shown by ollama ls. Run du -sh ~/.ollama/models again to see the space come back.

Prefer ollama rm over deleting files inside the folder by hand. It is the documented way, and it keeps Ollama's own list matching what is on disk.

How do I know which models are safe to remove?

Start with models you pulled to try once. A few checks help:

• Run ollama ps to see which models are loaded right now. If a model is running, stop it first with ollama stop <model>.
• Look at the date shown by ollama ls. An old date is a hint, not proof: you may still use that model every week.
• Ask whether you could download it again. A 4 GB model is a quick re-download. A 40 GB model on a slow connection is not.

Ollama has no Trash step. ollama rm removes the model, so the only undo is pulling it again.

What about Hugging Face and LM Studio models?

Ollama is only one place local models hide. The Hugging Face hub cache lives in ~/.cache/huggingface/hub (see where Hugging Face stores models on a Mac), and LM Studio keeps its downloads in ~/.lmstudio/models, according to the LM Studio docs. The same models can sit in more than one of them, which is easy to forget.

Does deleting a model break Ollama?

No. Ollama keeps working. The model is simply gone until you run ollama pull for it again.

A safer way: see the size and last use first

The terminal gives you names and sizes, but not a single view of everything, and nothing is reversible. Disk Goat is an offline Mac app we are building for exactly this: it lists every model and cache with its size and last-used date, never ticks a model for you, shows a dry run, and moves what you pick to the Trash instead of deleting it. It is not released yet, but you can click through a demo of it.