GPT-5.1 was released last night, and it barely mentioned any differences in reasoning abilities. Instead, it emphasized the ability to dynamically adjust computational power for tasks of varying difficulty. So, is the goal this time to save costs?