How to Delete or Remove Ollama Models (PowerShell, Terminal & Windows)

Running out of disk space is the number-one reason people look up how to remove Ollama models — and the cleanup genuinely is mostly one command. The catch is that a few things look like Ollama “deleting” a model but silently leave gigabytes of orphaned blobs behind. This guide covers the reliable ways to remove one model, nuke them all, clean up leftovers, and fully uninstall Ollama, on Windows, macOS, and Linux — with the exact paths as they actually appear on each OS.

Step 1: See what’s installed

Before deleting, list what you have and roughly how much space each takes. This works in PowerShell, Terminal, or any shell:

ollama list

The NAME column is what you’ll copy. It can include a tag (e.g. llama3.1:8b), and the match is exact — so copy rather than retype.

Step 2: Delete a single model

Use ollama rm with the full name:

ollama rm llama3.1

If it has a tag, include it to target that exact variant:

ollama rm llama3.1:8b

The command returns quickly and frees the space immediately. Confirm with ollama list — the model should be gone. If you get “Model Not Found,” the name doesn’t match what’s in the list (case-sensitive on some systems); copy it straight from ollama list.

Step 3: Delete every model at once

Ollama has no built-in “remove all” flag, so loop over the list. The loop differs per shell.

Windows (PowerShell)

ollama list | Select-Object -Skip 1 | ForEach-Object { ollama rm ($_ -split '\s+')[0] }

macOS / Linux (bash)

ollama list | tail -n +2 | awk '{print $1}' | xargs -I {} ollama rm {}

Both skip the header, extract each name, and run ollama rm. This frees several gigabytes per model, so check your free disk afterwards.

If you’re actually hitting VRAM pressure (model too big for your GPU) rather than disk pressure, deleting models isn’t the right lever — pick a smaller quantization instead. The Ollama GPU and memory errors guide explains the sizing math.

Where Ollama keeps the files

Inside there are two subfolders: manifests (small metadata that says what a model is) and blobs (the weight files that actually eat gigabytes). Knowing this matters for the next step: most “it’s not freeing space” confusion traces back to leftover blobs.

Manual cleanup of leftover blobs

Interrupted downloads or manual edits can leave orphaned blobs that ollama rm doesn’t touch. Because several models can share a single blob, never delete individual files while models are still installed — you can silently corrupt another model. The safe path is: remove models via the CLI first, then, if you still need to reclaim space, delete the whole models folder.

Windows (PowerShell)

Remove-Item -Recurse -Force "$env:USERPROFILE\.ollama\models"

macOS / Linux (bash)

rm -rf ~/.ollama/models

How to fully uninstall Ollama

Removing models is only half of a full uninstall — the app and its launcher are the other half. Leaving the data folder behind keeps gigabytes on your disk.

Windows

Settings → Apps → uninstall Ollama. Then, from PowerShell, remove the leftover data and the app folder if the uninstaller left them:

Remove-Item -Recurse -Force "$env:USERPROFILE\.ollama"
Remove-Item -Recurse -Force "$env:LOCALAPPDATA\Programs\Ollama"

Restart to clear any lingering background process.

macOS

Drag the app from Applications to the Trash, or if you installed via Homebrew:

brew uninstall ollama
rm -rf ~/.ollama

Linux

Stop and disable the service first, then remove the unit, binary, user data, and service user:

sudo systemctl stop ollama
sudo systemctl disable ollama
sudo rm /etc/systemd/system/ollama.service
sudo rm $(which ollama)
sudo rm -rf /usr/share/ollama
sudo userdel ollama
sudo systemctl daemon-reload

Remove any user-level data too with rm -rf ~/.ollama.

Conclusion

Deleting an Ollama model is a two-command job: ollama list to see what you have, then ollama rm <model> to remove it. For bulk cleanup, loop over the list or, after CLI removals, delete the models folder directly. If clean disk space is your real goal, remember the .ollama\models / .ollama/models data folder is where the gigabytes live, and it’s only fully gone once you remove it too.

If you’re doing a fresh, properly configured install afterward, the Ollama Complete Guide walks through a clean setup. Keep only the models you actually use, and your local AI rig stays fast and light.