What is a foundation model?
Quick answer
A foundation model is a large AI model trained on a huge, broad mix of data so it can be adapted to many jobs: writing, summarizing, answering questions, reading images. GPT, Claude, Gemini and Llama are foundation models, and most AI products are built on top of one.
Last updated
Updated · By Robert Breen
Why it matters for a small business
When a vendor says its tool "uses AI," it usually means it sends your request to a foundation model from OpenAI, Anthropic, Google or another lab, wrapped in its own instructions and screens. Knowing that helps you ask the right questions: which model is underneath, where your data goes, and what happens when the vendor switches models.
It also explains why one subscription can do so many different things. The model was not built for your invoice emails or your social posts. It learned general patterns of language, and a system prompt, some examples or your own documents point it at your job. The same foundation can be a marketing helper in the morning and a bookkeeping helper in the afternoon.
In a real lesson: n8n AI Agent Tutorial: Save Social Media Ideas to Google Sheets
In the AI Social Media Idea Agent lesson, you build an n8n agent for BrightPath Marketing, a made-up digital marketing agency. You add an OpenAI Chat Model node and pick gpt-5-mini. That model is the foundation. Everything that makes the agent about BrightPath comes from the system prompt.
The prompt says the agency helps "small and midsize businesses with SEO, paid advertising, social media marketing, and website design," sets a brand voice, and asks for a Platform, Post Topic, Hook, Short Description and Suggested Call to Action for each idea. It also says "Do not invent statistics."
The Deal Follow-Up Agent lesson uses the same kind of model node for Northwind Scheduling Software, a different made-up company, with a different system prompt and a different sheet. Same foundation, different job. That is the whole idea in two lessons.
Try this lesson free or read the step-by-step guide.
Common confusions
Foundation model vs AI product
ChatGPT is a product: an app with chat history, file uploads and settings. The model underneath is the foundation. Several products can share one foundation model, and one product can offer several.
Foundation model vs fine-tuned model
A fine-tuned model is a foundation model given extra training on a narrower set of examples. Most office use never needs that step. Good instructions and your own reference material usually get you there.
Tips
- Ask any AI vendor which foundation model their tool uses and what their contract says about your data.
- When a vendor updates its model, rerun a few of your saved test prompts. Output can change even when your prompt does not.
- Put your business facts in the prompt or a knowledge file. The foundation model does not know your prices or policies.
Related terms
More AI basics terms
Where you use it: free lessons
- n8n AI Agent Tutorial: Save Social Media Ideas to Google Sheets (n8n, 12 min)
- Build an n8n AI Agent That Writes Deal Follow-Ups (n8n, 12 min)
- Build an n8n AI Agent That Plans Your Content Calendar (n8n, 12 min)
Frequently asked questions
- Where did the term foundation model come from?
- Researchers at Stanford popularized it in 2021 to describe large models that many other systems are built on. You will also hear frontier model, which usually means the most capable foundation models available at a given time.
- Do I need to pick a foundation model myself?
- In ChatGPT or Copilot the product picks for you, sometimes with a menu. In tools like n8n you choose it in a model node, as the lessons show, so it helps to know the trade-off between cost, speed and capability.