This is key info from the article for me: > ------------------------------- "reasoning": {"summary": "auto"} }' Here’s the response from that API call. h

This is key info from the article for me: > ------------------------------- "reasoning": {"summary": "auto"} }' Here’s the response from that API call. https://gist.github.com/simonw/1d1013ba059af76461153722005a039d https://gist.github.com/simonw/1d1013ba059af76461153722005a0... Without that option the API will often provide a lengthy delay while the model burns through thinking tokens until you start getting back visible tokens for the final response.