The point at which a model's performance stabilizes during training and stops significantly improving, indicating the training process is complete.