Ollama
Ollama model storage on Mac: model blobs, tags, and stale pulls
Local AI workflows can consume tens or hundreds of gigabytes quickly. Ollama stores model blobs and tags in dedicated paths that are easy to overlook during normal cleanup.
Updated:
Read time: 6 min
Why this silently grows
Large models are downloaded as layered blobs. Testing multiple versions or forks can keep old blobs and tags resident even after workflows change. Because these files are not in obvious project directories, growth often goes unnoticed until storage pressure becomes urgent.
How Free Mac Space finds it
Free Mac Space surfaces Ollama-related directories such as ~/.ollama/models, ~/.ollama, /usr/local/share/ollama, and /opt/homebrew/var/lib/ollama, then ranks them by size so you can identify stale model stores quickly.
How cleanup is handled
Review model directories by size and modified time, then remove stale pulls in a controlled order. Cleanup keeps Trash-first behavior and path validation before any action is executed.
Safety boundary
Model deletion can disrupt active local agents and scripts. Keep currently deployed models and remove only versions that are no longer referenced by your workflows.
Paths covered in Free Mac Space
- ~/.ollama/models
- ~/.ollama
- /usr/local/share/ollama
- /opt/homebrew/var/lib/ollama
Recommended monthly check
- Identify the largest model folders and check their last use.
- Keep models required by active local automation or demos.
- Re-scan after cleanup to verify reclaimed capacity before pulling new models.
Step-by-step workflow
1. Identify why Ollama storage keeps growing
Large models are downloaded as layered blobs. Testing multiple versions or forks can keep old blobs and tags resident even after workflows change. Because these files are not in obvious project directories, growth often goes unnoticed until storage pressure becomes urgent.
2. Inspect the highest-impact paths first
Free Mac Space surfaces Ollama-related directories such as ~/.ollama/models, ~/.ollama, /usr/local/share/ollama, and /opt/homebrew/var/lib/ollama, then ranks them by size so you can identify stale model stores quickly. Priority paths: ~/.ollama/models, ~/.ollama, /usr/local/share/ollama, /opt/homebrew/var/lib/ollama.
3. Confirm the safety boundary before acting
Model deletion can disrupt active local agents and scripts. Keep currently deployed models and remove only versions that are no longer referenced by your workflows.
4. Apply a review-first cleanup workflow
Review model directories by size and modified time, then remove stale pulls in a controlled order. Cleanup keeps Trash-first behavior and path validation before any action is executed.
5. Monthly validation step 1
Identify the largest model folders and check their last use.
6. Monthly validation step 2
Keep models required by active local automation or demos.
7. Monthly validation step 3
Re-scan after cleanup to verify reclaimed capacity before pulling new models.
Frequently asked questions
What hidden storage sources are covered for Ollama?
Primary sources include Model blobs, duplicated tags, pulled model history. Large models are downloaded as layered blobs. Testing multiple versions or forks can keep old blobs and tags resident even after workflows change. Because these files are not in obvious project directories, growth often goes unnoticed until storage pressure becomes urgent.
Which macOS paths should I inspect first?
Start with: ~/.ollama/models, ~/.ollama, /usr/local/share/ollama, /opt/homebrew/var/lib/ollama. Free Mac Space surfaces Ollama-related directories such as ~/.ollama/models, ~/.ollama, /usr/local/share/ollama, and /opt/homebrew/var/lib/ollama, then ranks them by size so you can identify stale model stores quickly.
How can I reduce this storage safely?
Review model directories by size and modified time, then remove stale pulls in a controlled order. Cleanup keeps Trash-first behavior and path validation before any action is executed. Model deletion can disrupt active local agents and scripts. Keep currently deployed models and remove only versions that are no longer referenced by your workflows.
What should the monthly review checklist look like?
Identify the largest model folders and check their last use. Keep models required by active local automation or demos. Re-scan after cleanup to verify reclaimed capacity before pulling new models.