I created this page to apply data envelopment analysis to find the Pareto frontier of cost and benchmark scores. The goal is to identify the cheapest model or a theoretical blend of models applied to tasks that meets the same benchmark scores as your selected model. This is based on a 2022 paper I coauthored. Benchmark sources include Artificial Analysis, ToneBench, Hermes3000 and Slidely, from September–October 2026.