Not every AI prompt deserves multiple seconds of thinking: how Meta is teaching models to prioritize

Let models explore different solutions and they will find optimal solutions to properly allocate inference budget to AI reasoning problems.