4.6 was the last model that let you disable adaptive thinking and set max thinking token budget. I liked having that available, and still use it sometimes.