National

Small Language Models Study: 94% Task Efficiency vs Large Models (MIT Research 2025)