How do you run a model locally on a Mac: Ollama, LM Studio, MLX
Essence
A local model is an open-weight language model running right on your Mac: no internet, no API key, no sending code to third parties. Apple silicon Macs are well suited for this thanks to unified memory: the GPU can use most of the RAM, so 20–30 billion parameter models run on a laptop with 32 GB.
The price is quality and speed. A local 8–30B model is noticeably behind the cloud flagships on complex agentic tasks, but it handles code explanations, drafts, translation, classification and experiments very well.
The rest of this answer, with code and interview follow-ups, is in the DevPrep app.
Read the full answer in DevPrep