I'm thinking a Mini size machine with an Apple supported LLM, 4e cores and a huge number of gpu cores. This would be for inference with voice response, etc. Thoughts?