INTELLIGENCE PER WATT: MEASURING INTELLIGENCE EFFICIENCY OF LOCAL AI

Summary from the TLDR AI newsletter

Frontier models today are served through massive, centralized cloud computing services that are straining electric grids. The performance gap between these large models and much smaller models, the kind that can run on a phone or laptop, is shrinking.