A model that has already been trained on large amounts of text data before being released or fine-tuned for specific tasks.