I always expose reasoning traces. How else can you seriously debug?
Those traces are just summaries of the reasoning.
The closed models aren't giving you the real thinking traces, though.
Those traces are just summaries of the reasoning.