@BennettBuhner Yeah to me it feels the same, and realistically they don’t really have a way to make a non-reasoning model just reason, so they must send the request off to another reasoning model each time (I assume), so idk why it’d be different. I wouldn’t be surprised if they update it soon