Identify the best AI models to run locally on specific hardware configurations. The site provides a database of performance trials featuring metrics such as decode speed, prefill time, and peak memory usage, enabling users to compare models based on real-world data from various devices.