How well a model performs relative to the computational resources (processing power and memory) required to run it.