The amount of processing power and memory available to run a model, which determines how much computation can be performed.