Question 1

What's the difference between a foundation model and a large language model (LLM)?

Accepted Answer

A foundation model is a broad term for a large model trained on diverse data that can be adapted to many tasks (text, images, audio, or multiple). An LLM is a type of foundation model focused on language tasks like writing, summarizing, and coding. So: many LLMs are foundation models, but not all foundation models are LLMs (some are vision or multimodal).

Question 2

When should I use a foundation model instead of training my own model?

Accepted Answer

Use a foundation model when you need strong general performance quickly, you don’t have enough labeled data to train a high-quality model from scratch, or you want to support many tasks (chat, summarization, extraction, classification) with one base capability. Train your own model from scratch only when you have very specialized requirements, large proprietary datasets, and the budget and expertise to manage training, evaluation, and ongoing updates.

Question 3

How much does a foundation model cost?

Accepted Answer

Costs usually depend on (1) tokens processed (input + output) for text models, (2) image/audio units for multimodal models, (3) model size and latency tier, and (4) whether you use on-demand APIs or provisioned/committed capacity. Additional costs often include vector databases for RAG, data storage, content filtering/safety services, and monitoring. Fine-tuning or customization (if used) adds one-time training costs plus ongoing hosting/inference costs.

Foundation Model

Definition

Real-World Example

Related Terms

Cloud Provider Equivalencies

Explore More Cloud Computing Terms