I run a local LLM server on an M4 Pro Mac mini — here's my setup

My local model setup on an M4 Pro Mac Mini

A developer details how they run a local LLM server on an M4 Pro Mac mini with 48GB RAM, using Qwen3.6-35B-A3B and Gemma-4-E4B models via oMLX, with Tailscale for network access. They explain the benefits: cost predictability, latency, offline capability, no rate limits, and data privacy. The post also demystifies MoE model memory requirements and shows how to swap models easily.