Output o3 chain of thought in agent mode

Feature Requests
1

It would be useful to output the 03-mini (and other thinking models) chain of though reasoning in the agent mode so I can see if the model has correctly understood the prompt request.

Right now it’s a bit unsettling how o3-mini just makes code changes with no explanation. I would trust it more if I could see it’s reasoning. This is especially important since 03-mini is so much more powerful than most other models right now for doing heavy lift analysis and complicated changes.

2 Likes
2

Thanks for the suggestion about showing chain of thought reasoning in agent mode - totally get why seeing the model’s thought process would help build trust. Will pass this feedback along to the team!

3

I second this as well; the model is cost-effective and powerful but I’d really love to see the reasoning behind its actions.