It seems the recitation problem has been fixed with the latest models. In my tests, the answer generation no longer stops prematurely. Before I resume a project that was on hold due to this issue, I'm gathering feedback from other users about their experiences with this and the new models. Are you still having this problem?
