What is a foundation model? Models, assistants and plugins
Understand a foundation model in plain language, how it differs from an LLM, and where AI assistants and development plugins fit in a task.
Why is it called a foundation model?
A foundation model is trained on broad data and can be used or adapted for many downstream tasks. Think of it as a general capability base that different applications can build on, rather than a model trained for only one job.
How is it different from an LLM?
A large language model, or LLM, is one kind of foundation model focused on language. The broader category can also involve images, audio and other data. A foundation model can support classification or analysis as well as content generation.
Model, assistant or plugin?
A model supplies capabilities such as understanding and generation. A development assistant is the workspace where you submit tasks and review changes or actions. Plugins add instructions, resources or tool connections to a compatible assistant; installing one does not mean changing the underlying model.
To check a web page, for example, you need an available assistant and a suitable browser connection. Naming a model alone does not establish that this workflow is ready.
Do you need to train a model yourself?
For using an assistant or a listed plugin, start with access, task capabilities and setup requirements. Model training is not a first step. When building your own application, decide how to connect or adapt a model around the task and available information.
- Describe a task such as understanding code, retrieving documentation or interacting with a page.
- Check the chosen tool’s inputs, external connections and working environment.
- Try a task you can verify, review the result and then continue.
Sources and next steps
Sources checked on 2026-10-12. This explains concepts and selection, without accuracy, speed or price rankings.