Practical guides for developers running AI models locally on Mac.
Local AI cleanup is not the same as deleting random large files. Models, caches, project memory, and shared storage all need different treatment. Here is a safer approach.
LM Studio makes it easy to download models — which means it is easy to accumulate gigabytes of GGUF files you no longer use. Here is what to check before cleaning up.
AI coding assistants like Windsurf and Codeium create local cache and index files. Some are rebuildable. Some should be reviewed carefully. Here is what to consider.
Cursor creates workspace indexes and AI context folders as you work. They can pile up across every project you have opened. Here is what to understand before cleaning them up.
Claude Code and similar AI coding tools keep local project memory and context. That is not always junk — deleting it blindly can remove useful working context. Here is what to know.
Running agents locally with open-source LLMs feels private and powerful — but the storage footprint is easy to underestimate. Here is what is actually accumulating.
Local AI agents do not only use models. They create caches, indexes, logs, embeddings, and project memory. Here is what is actually accumulating — and why it is harder to clean than model files.
You may have downloaded models through Python, Transformers, Whisper, or Sentence Transformers — and never realised Hugging Face kept them locally. Here is what to know before cleaning up.
Downloading a few local models can fill your SSD faster than expected. Here is what is actually happening — and why it is harder to clean up than it looks.
Ollama models are powerful, but they quietly take up gigabytes inside hidden folders most users never open. Here is what to know before you start deleting.