The amount of memory and computational resources required to run a model, determined primarily by its size and architecture.