cleanyourmac.space
AI & ML•4 min read•Aug 20, 2026

Where Local AI Models Hide on macOS: Cleaning Ollama & LM Studio GGUFs

Dr. Aaron Chen
AI Infrastructure Specialist

Quick answer

Run `ollama list` to see installed models and `ollama rm <model>` to delete one; that frees the blobs in ~/.ollama/models. LM Studio keeps models in ~/.lmstudio/models (older versions: ~/.cache/lm-studio/models), and Hugging Face downloads live in ~/.cache/huggingface/hub — delete the model folders you no longer use.

Table of Contents

1. The Footprint of Local LLMs

With Unified Memory architectures on Apple Silicon (M1/M2/M3/M4 Pro & Max), running AI models locally is fast and private. However, a single 7B model requires ~4.5 GB, 14B requires ~9 GB, and 70B models exceed 40 GB. Testing 4–5 models can easily consume half of a 256 GB or 512 GB MacBook SSD.

2. Default Storage Paths on macOS

  • Ollama Blobs: ~/.ollama/models/blobs/ (Content-addressed SHA-256 binary blobs)
  • Ollama Manifests: ~/.ollama/models/manifests/
  • LM Studio Cache: ~/.cache/lm-studio/models/
  • HuggingFace Hub: ~/.cache/huggingface/hub/

3. Listing & Deleting Ollama Models

# View all installed models with exact sizes
ollama list

# Delete an unneeded large model
ollama rm llama3.3:70b

Skip the Terminal: see it all in one scan

CleanYourMac Space maps every cache, build folder and model file by size, then cleans only what you pick — everything goes to the Trash first, so it's always recoverable.

Download free for MacOllama & LLM Model Cleaner

Related Technical Guides

Storage Guide

Emergency Guide: How to Fix "Your Disk is Almost Full" on Mac

Read
Storage Guide

Best Mac Disk Space Analyzers in 2026: Visual Treemap vs List View

Read
Developer

How to Safely Clear 50+ GB of Xcode DerivedData and Archives in 2026

Read