A model trained on large amounts of text data to predict and generate language before being adapted for specific applications.