>I can't tell from the Gist alone

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