I have a Qwen model running on a Mac Mini with Ollama - it’s reasonable for Home Assistant, responding to iMessages and similar.