llama is a text prediction model similar to GPT-2, and the version of GPT-3 that has not been fine tuned yet.

It is also possible to run fine tuned versions like vicuna with this. I think. Those versions are more focused on answering questions.