download a model and run it. There are many tutorials online. Also it uses llama.cpp, which basically means that you must use models with a .gguf file format. This is the most common format nowadays and has very good support. As for what model to run, it depends on the memory of ...