Closed models just don't show this thinking process directly to the user
Depends on harness.
No, I mean they don't show it at all in their API to prevent distillation. The reasoning is encrypted and only accessible by their server. They sometimes show a summary of the reasoning, but it's not the same as the actual contents of the reasoning.
Thanks I didn't know that.
From a recent DeepSeek session:
"Wait, what am I? I am claude, or something similar"