A foundation model is a large model pre-trained on broad data that serves as the base for many downstream uses: adapted via prompting, retrieval, or fine-tuning
A large model pre-trained on broad data that serves as the base for many downstream uses, adapted via prompting, retrieval, or fine-tuning rather than built per task. LLMs, vision-language models, and multimodal systems are all foundation models.
They reorganized it: a few organizations train foundations, and everyone else builds on top. Understanding what the foundation does and does not know, and where adaptation is needed, is the starting point of applied AI architecture.