A snapshot of a model saved during training before it has finished learning, allowing you to use or study the model at an intermediate stage rather than waiting for full completion.