I thought I had heard that the integrated RAM/VRAM makes local LLMs fairly quick on a RAM-maxxed Mac Mini.