We have lots of users that mix in local models for some tasks or fully run on local models. (that requires beefy hardware so fewer folks do that so far)

→ View original post on X — @steipete