Apple shows how much faster the M5 runs local LLMs compared to the M4