we have very efficient models, especially for their capability level happy codexing

→ View original post on X — @sama,