Reasoning Traces for O models

Hi,

Like deep seek r1, is there a possibility to log the reasoning traces through the APIs. In the app interface I see summary of the reasoning trace (not full reasoning trace) for o3-mini

There is a future possibility, but not currently.

The only model that will return the reasoning summary is computer-use-preview.

This is despite o1 models returning an output list item for this summary before the item with the model output, breaking naive handling of response.output. It is not turned on, and empty.

Even when I have reasoning set to concise, nothing is ever output from computer-use-preview unless I’m missing something