A pre-trained model is an AI system that has already been trained on large datasets and can be used directly or adapted through fine-tuning for specific tasks. Rather than building a model from scratch, organizations start with a pre-trained foundation and adapt it to their needs.
They provide a strong baseline of language understanding and general knowledge that would take enormous resources to recreate from scratch. Fine-tuning a pre-trained model for a specific domain is dramatically faster and cheaper than training a specialized model from zero — and the resulting quality is typically better than what most organizations could achieve on their own.
