The amount of memory and computational resources required to run a model, with smaller footprints being more efficient.