It's reasoning.mode, next to reasoning.effort in the responses API. Set to the string "pro" for pro mode. Sorry, I don't use first-party harnesses so have no idea what it maps to there.