We do not plan to make Mythos Preview generally available. Our goal is to deploy Mythos-class models safely at scale, but first we need safeguards that reliably block their most dangerous outputs.
We’ll begin testing those safeguards with an upcoming Claude Opus model.