ParentFull threadmutex_man·You could run it on a single high end GPU. I can run llama2's models ,(except 70b) on my 4080.View on HN