Foundation Model
Working knowledgeCore AI Concepts
A foundation model is a large machine learning model trained on vast, multi-modal datasets using self-supervised learning. It arrives able to reason, read and write across general subjects without further training, which is what makes it a base rather than a finished product. One foundation model can be adapted to hundreds of downstream applications through simple prompting or fine-tuning.
In practice
Choosing a foundation model is an architecture decision, not a procurement one: it sets what your applications can do, what they cost to run and how hard it is to change course later. Ask what would have to be rebuilt if you moved to a different provider.