Ollama Local LLM Models Cache Folder (~/.ollama/models)

Bottom Line: It is 100% safe to delete unused Ollama models. Running ollama rm <model> or clearing ~/.ollama/models/blobs will NOT delete your Python AI code, prompt histories, or application settings.

What is the Ollama Models Cache Folder?

Located at ~/.ollama/models on Linux/macOS (and %USERPROFILE%\.ollama\models on Windows), this folder is managed by Ollama (the local Large Language Model runner).

When you run ollama run llama3 or test open-source AI models (Mistral, Phi-3, Gemma, CodeLlama), Ollama downloads quantized GGUF tensor model files into models/blobs. Because 7B, 13B, and 70B parameter models weigh 4 GB to 40 GB each, experimenting with multiple LLMs hoards 30 GB to 150 GB+ of SSD storage.

Can You Delete Ollama Models?

✅ Yes, it is completely safe to delete.

• Zero code loss: Your API integration scripts, LangChain/LlamaIndex code, and chat UIs remain 100% safe.
• Auto-downloaded: Running ollama run <model> in the future will simply re-fetch the model blob from Ollama Library.

How to Safely Clean Up Ollama Models

Method 1: Remove Models via Ollama CLI (Recommended)

1. Open your terminal.
2. List installed local models and sizes:

   ollama list

3. Remove models you no longer need:

   ollama rm llama2
   ollama rm mistral

Method 2: Purge Orphan Model Blobs

# macOS / Linux
rm -rf ~/.ollama/models/blobs/*

# Windows (PowerShell)
Remove-Item -Recurse -Force $env:USERPROFILE\.ollama\models\blobs\*

Related Guides

• Learn about deleting Hugging Face cache.
• How to clean up PyTorch model cache.
• Check if you can clean Docker Data.