and… 16K output tokens per request! much under- rated feature of the latest model.

→ View original post on X — @jiquanngiam