Running large language models locally requires shrinking their file size without destroying quality. This guide walks through the full llama.cpp pipeline: downloading a Hugging Face model, converting it to GGUF format, and quantizing it to Q4_K_M or other levels to fit consumer hardware.