April 2026 TLDR setup for Ollama + Gemma 4 26B on a Mac mini (Apple Silicon) — auto-start, preload, and keep-alive

Step-by-step guide to setting up Ollama with Gemma 4 26B on an Apple Silicon Mac mini. Covers installation via Homebrew, pulling and running the model, configuring auto-start at login, creating a launchd agent to preload the model into memory, and setting OLLAMA_KEEP_ALIVE to prevent model unloading. Also covers GPU acceleration via MLX, memory considerations (20GB usage on 24GB machines), and using the OpenAI-compatible local API at localhost:11434.