# Is it safe to delete Ollama models in ~/.ollama/models?

Is it safe to delete? · AI tools · Updated 21 November 2026. By Mohit Nippanikar, who makes SiftMac.

## The short answer

- **Yes, with `ollama rm`.** Each model you pulled is stored as layers in `~/.ollama/models`, and models can share layers. Ollama’s own command frees what only that model used; deleting files in Finder can break the models you keep.
- **What it costs you:** the model, until you pull it again, which means downloading it again. A model you built with `ollama create` comes back only if you still have its Modelfile.
- **Does it come back?** Only with `ollama pull`. Layers another model still uses are already on your Mac and don’t download again.
- SiftMac’s verdict: You decide

## 1. What it is

Ollama runs open-weight language models on your Mac. Every model you fetch with `ollama pull` or `ollama run` is downloaded once and kept until you remove it, so the folder only grows as you try new models.

A model isn’t a single file. Ollama splits it into two parts:

- `blobs/`: the layers, each file named after its SHA-256 digest (`sha256-…`). The weights are the big one; templates, parameters and licences are small.
- `manifests/`: one small file per model and tag, such as `registry.ollama.ai/library/<model>/<tag>`, listing the blobs that model uses.

Two tags of the same model, or a model you made from another with a Modelfile, can point to the same weights blob. There’s no folder per model to delete, which is why removing models is a job for Ollama itself.

## 2. Where it is

```
~/.ollama/models
```

That’s the default Ollama’s FAQ gives for macOS. If the `OLLAMA_MODELS` environment variable is set, models live in that folder instead; for the Ollama app, the FAQ says to set variables with `launchctl setenv` and restart the app. Moving them to another drive is covered in [moving Ollama and LM Studio models to an external drive](https://siftmac.com/guides/move-ollama-models-to-external-drive).

## 3. How big it gets

A model’s size depends on its parameter count and how it was quantized. `ollama list` shows each model with its size. Those sizes are per model, so a layer two models share is counted for both, and the column can add up to more than the folder really holds. For the real total:

```
ollama list
du -sh ~/.ollama/models
```

## 4. How to delete it safely

1. Check what’s loaded with `ollama ps`, and stop a model you’re about to remove with `ollama stop <model>`.
2. List your models with `ollama list` and pick the ones you no longer use.
3. Remove each one by its full name, tag included.

```
ollama rm <model>:<tag>
```

Ollama deletes the model’s manifest and any layers no other model refers to. Layers still used by another model stay, so that model keeps working.

Don’t delete files in `blobs` by hand: a blob may belong to several models, and a model whose layers are missing can’t load. To start again from nothing, remove every model with `ollama rm` and check that `ollama list` is empty.

Ollama also tidies `blobs` for you: when it starts, it removes blobs that no model refers to any more, such as parts of an interrupted download, unless the `OLLAMA_NOPRUNE` variable is set.

## 5. What happens afterwards

- The model disappears from `ollama list` and from any app that talks to Ollama, such as chat front ends or editor plugins set to use it.
- Using it again means `ollama pull <model>:<tag>`, or `ollama run`, which pulls first. Only the layers that aren’t already on your Mac download.
- Models you kept are untouched, including any that shared layers with the one you removed.

## 6. Where else local models hide

Ollama keeps its own copies. [LM Studio](https://siftmac.com/safe-to-delete/lm-studio-models) and the [Hugging Face cache](https://siftmac.com/safe-to-delete/hugging-face-cache) keep theirs, so a model you downloaded in all three takes its space three times. Check each one before deciding which copy to keep.

## Questions

### Can I delete the whole ~/.ollama folder?

You can, with Ollama quit, but it takes every model at once along with Ollama’s other files there. `ollama rm` lets you choose model by model.

### Why didn’t `ollama rm` free as much as `ollama list` showed?

The model shared layers with another one you kept. Those layers stay until the last model using them is removed.

### Does Ollama delete old models by itself?

No. Models stay until you remove them. At startup Ollama only clears out blobs that no model refers to.

### How do I get a model back?

Pull it again with `ollama pull <model>:<tag>`. You need an internet connection and the time to download whatever isn’t already on your Mac.

## How SiftMac handles it

Its AI files category lists each Ollama model by name and tag, sized by the layers only that model uses. Models are You decide and never ticked for you; when you choose one, SiftMac removes it with `ollama rm`, so layers other models share stay where they are.

Download: https://siftmac.com/SiftMac.dmg (scanning is free, and so is the first 2 GB of cleaning).

## Sources

- [Ollama FAQ: where models are stored, OLLAMA_MODELS](https://docs.ollama.com/faq)
- [Ollama: CLI reference](https://docs.ollama.com/cli)
- [Ollama on GitHub: environment settings (OLLAMA_MODELS, OLLAMA_NOPRUNE)](https://github.com/ollama/ollama/blob/main/envconfig/config.go)

## Related

- [Is it safe to delete LM Studio’s models folder?](https://siftmac.com/safe-to-delete/lm-studio-models)
- [Is it safe to delete ~/.cache/huggingface?](https://siftmac.com/safe-to-delete/hugging-face-cache)
- [How to move Ollama and LM Studio models to an external drive](https://siftmac.com/guides/move-ollama-models-to-external-drive)
- [What Claude Code, Cursor, Ollama and Hugging Face leave on your Mac](https://siftmac.com/guides/ai-files-on-mac)
- [What fills a Mac that runs local models and AI coding tools](https://siftmac.com/for/ai-developers)

Canonical: https://siftmac.com/safe-to-delete/ollama-models
