Why Don't You Just Use GPT or Claude?

CPU inference models are also on their way which will help immensely with running LLMs locally without having to spend sheer $$ on high end GPU's.

https://github.com/microsoft/BitNet?tab=readme-ov-file