Mark Zuckerberg made a good point in his Lex Friedman podcast referring to LLAMA, that he foresees not one master model but a bunch of speciality purpose smalle

Mark Zuckerberg made a good point in his Lex Friedman podcast referring to LLAMA, that he foresees not one master model but a bunch of speciality purpose smaller models built on base models with fewer parameters