There are lots of reasons to run LLMs locally, from privacy concerns to just wanting to futz about with the technology, but if you want to run the big models, the standard logic is that you need bi…