MacaroomWhere Ollama and LM Studio keep models on Mac, and how to delete them
Every local AI model you try is often several gigabytes, and they stay on disk after you move on. Here's where they are and how to remove them cleanly.
Where the models are
| Tool | Folder | Best way to remove |
|---|---|---|
| Ollama | ~/ | ollama rm followed by the model name |
| LM Studio | ~/ | Delete it from LM Studio's model list |
| Hugging Face | ~/ | Delete the model's folder (named models--org--name) |
Ollama: use ollama rm
Run ollama list to see your models and their sizes, then remove one with ollama rm and the model's name, for example ollama rm llama3.2.
Don't delete files in ~/.ollama/models/blobs by hand. Ollama stores models as shared pieces with separate lists of which pieces each model needs, so deleting pieces directly can break other models and leave Ollama thinking a model is still there.
If you want to keep models but not on your Mac's internal disk, Ollama can store them elsewhere: set the OLLAMA_MODELS environment variable to a folder on another drive.
What you lose
Nothing but the download. A deleted model has to be downloaded again before you can use it, often several gigabytes, so keep the ones you use regularly.
How Macaroom handles it
Macaroom finds models from Ollama, LM Studio and Hugging Face, plus loose model files elsewhere, and marks them Rebuildable. LM Studio and Hugging Face models are listed one at a time. Ollama's models show as one item, with the ollama rm command for removing a single model. It also names the hidden folders AI coding tools keep, such as Claude Code, Codex and Gemini CLI, and explains which to keep.
Related guides
See it on your own Mac
Macaroom scans your Mac, names what each chunk of storage is, and marks what's safe to clear. Anything you remove goes to the Trash, so you can put it back.
Get Macaroom