Foundation Model
A foundation model is a large model trained on broad data at scale — typically self-supervised — that can be adapted to many downstream tasks via prompting, fine-tuning, or retrieval, instead of being trained task-by-task.
Full definition
The term was coined by the Stanford CRFM in 2021 to describe a paradigm shift: instead of training one model per task, train one general-purpose model on broad data and adapt it. Foundation models include LLMs (text), diffusion models (image), and increasingly multimodal models. They are the building block of the modern AI stack — a small number of foundation models, fine-tuned and orchestrated, now powers most enterprise AI.
Why it matters
Foundation models concentrate AI capability — and risk — in a handful of models from a handful of providers. This shifts enterprise procurement from "build a model" to "select a foundation model and adapt it," changing skills required, cost structure, and governance posture.
Example
A healthcare network selects GPT-4o as its foundation model, then fine-tunes a private copy on de-identified clinical notes for triage, prescription summarization, and patient-instruction generation — three different applications, one underlying model.
Related terms
- Large Language Model (LLM)A Large Language Model (LLM) is a deep neural network — almost always a transformer — trained on hundreds of billions to trillions of words to predict the next token, and to generate, summarize, translate, or reason over text.
- Fine-tuningFine-tuning is the process of further training a pretrained model on a smaller, task-specific dataset so it specializes in a particular style, domain, or behavior — without retraining from scratch.
- Multimodal ModelA multimodal model is an AI model that natively understands or generates more than one type of input — typically text plus images, audio, or video — within a single network rather than via separate task-specific models.
- Generative AIGenerative AI is a class of AI systems that produce new content — text, images, code, audio, or video — by learning the distribution of their training data and sampling from it, rather than classifying or predicting from existing inputs.
- TransformerThe transformer is a neural-network architecture built around the self-attention mechanism that has become the dominant model design for language, vision, audio, and multimodal AI since 2017.
Source & further reading
Primary source: Bommasani et al. — "On the Opportunities and Risks of Foundation Models" (Stanford CRFM) (2021).
Citation policy: this entry is part of the AIDOLS AI Implementation Glossary and may be quoted for research, journalism, and education with attribution to aidolsgroup.com/es/glossary/foundation-model/.