Can You Run LLMs Locally Without a GPU? I Tested 8 Models on Linux It's FOSS