Wow, this makes running LLMs locally significantly more approachable. Very cool!