In case you're not aware, Codex lets you set the reasoning effort in the model picker or with codex --config 'model_reasoning_effort="high"'.