How to delete Ollama models on Mac

Part of our guide to why System Data is so large on Mac · Updated September 2026 · 6 min read

Every model you ollama pull stays on disk until you remove it. Small models are 2–5 GB, 30B-class models around 20 GB, and 70B models 40 GB or more, so a few experiments can quietly take 100 GB. Here's how to see what you have, remove what you don't use, and why deleting files by hand is a bad idea.

1. List your models and their sizes

ollama list

This shows each model's name and tag, ID, size and when it was last modified (pulled or updated). "Modified" is not "last used": a model you pulled a year ago and run daily still shows the old date.

To see which models are loaded in memory right now:

ollama ps

2. Remove a model

ollama rm llama3.3:70b

Use the exact name:tag from ollama list. You can remove several at once: ollama rm model-a model-b. Pull any of them again with ollama pull; nothing about your setup changes except the download.

If a model is loaded, stop it first with ollama stop name:tag.

Where Ollama stores models on macOS

By default, models live in ~/.ollama/models, unless you've set the OLLAMA_MODELS environment variable. Inside are two folders:

manifests/: one small file per model and tag, listing the layers it uses.
blobs/: the actual weights and other layers, named by their SHA-256 hash (sha256-…).

du -sh ~/.ollama/models

Don't delete blobs by hand. Models built from the same base share layers, so removing a blob can break a model you still use, and Ollama will still list the model while its weights are gone. ollama rm only deletes layers no other model needs.

Storage Cleaner's AI Models list with Ollama models such as llama3.3:70b and qwen2.5-coder:32b, their sizes and when each was last used — "Last used" rather than "last pulled" is what tells you which models you can remove. (Storage Cleaner, demo data.)

How much space will I actually get back?

Because of shared layers, a model's listed size isn't always what you'll reclaim. If two models share a 4 GB base, removing one frees less than its listed size, and removing the second frees the rest. Check before and after with du -sh ~/.ollama/models.

Move models to an external drive instead

If you'd rather keep a large collection, store models on an external SSD. For the Ollama Mac app, set the variable for GUI apps and restart Ollama:

launchctl setenv OLLAMA_MODELS "/Volumes/AI/ollama-models"

If you run ollama serve from a terminal instead, export OLLAMA_MODELS in ~/.zshrc. Existing models stay in the old folder: move the contents of ~/.ollama/models to the new location (quit Ollama first), or pull them again. launchctl setenv doesn't survive a restart, so re-run it after rebooting, or add it to a login script.

Removing Ollama completely

Dragging Ollama to the Trash leaves the models behind. To remove everything, quit Ollama, delete the app, then delete ~/.ollama. That deletes every model and your Ollama settings.

Rather not do this by hand? Storage Cleaner finds every Ollama model in one scan, counts only the space you'd really get back, shows when you last used each one, and removes them safely with Undo.

Download free

Frequently asked

Does quitting Ollama free the space? No. Quitting frees memory, not disk. Only removing models frees disk space.

Why is ~/.ollama bigger than all my models added up? Usually it isn't. ollama list counts shared layers once per model, so the list can add up to more than the folder. If the folder is bigger, an interrupted download may have left partial blobs; pulling that model again or removing it cleans them up.

Do LM Studio or Hugging Face share Ollama's models? No, each keeps its own copies. See where LM Studio stores models and the Hugging Face cache.

More guides: Ollama model sizes in GB · How much disk space local LLMs need.