Just find a local LLM model according to your VRAM and you can have as many choices as you would like, at least for a few pages
It's actually pretty cool if you have the right expectations