Last checked: 6 November 2026
In short: An AI model is the trained part of an AI system: a large file of numbers that holds the patterns learned from data. Software sends it an input, and the model returns a prediction, an answer or new content.
Table of Contents
How it works
- Training. A learning algorithm runs over a large data set and adjusts millions or billions of numbers called parameters.
- Saving. Those numbers are stored as the model, which can be copied and run on other computers.
- Running. An app sends the model a question or data, and the model returns its output. This step is called inference.
Where you’ll meet it
ChatGPT, Claude, Gemini and Mistral’s Le Chat are apps. Each one runs on one or more AI models behind the scenes.
What it means for your business
- Models are swappable. Many tools let you switch between models, so you can compare cost and quality.
- Size affects price. Larger models usually give better answers but cost more for each request.
- Where it runs matters. Some models can run on your own servers, which keeps data inside your business.
Watch out for
- Version changes. Suppliers update or retire models, and answers can change overnight.
- Licence terms. Free or open models may limit commercial use. Read the licence.
Example
Mistral AI. In September 2025 the Paris model maker Mistral AI raised €1.7 billion in a funding round led by ASML, the Dutch chip equipment maker. Mistral said the round valued it at €11.7 billion after the investment. ASML put in about €1.3 billion, becoming its largest shareholder with around 11%, and the two firms agreed to work together on AI for ASML’s products.
AI model vs algorithm
| AI model | Algorithm | |
|---|---|---|
| What it is | Learned patterns saved as numbers | A set of step-by-step instructions |
| Made by | Training on data | People, or a learning process |
| Changes when | It is retrained on new data | Someone rewrites the steps |
Frequently asked questions
Is ChatGPT an AI model?
No. ChatGPT is an app. It sends your messages to AI models made by its supplier and adds features such as memory, web search and file uploads.
Should a business build its own AI model?
Rarely. Training a large model costs millions. Most firms rent an existing model and add their own documents and instructions, which is faster and far cheaper.
How often do AI models change?
Often. Leading suppliers release new versions every few months and retire older ones. Retest the tasks you depend on whenever your supplier switches models.