AI model: the engine behind the app
An AI model is what you get after training an AI: a very large set of numbers that turns a request into an answer. The app you use (ChatGPT, Gemini, Claude…) is just the bodywork; the model is the engine.
Data as of October 22, 2026
Key figures
| Level | Beginner |
|---|
The engine, not the car
When you message an AI assistant, your text is sent to an AI model. The model is a big file of numbers, its parameters, set during training from training data. For each request, it runs calculations with those numbers and produces an output: text, an image, a transcript.
The app handles everything else: the chat window, your history, the button to attach a file, sometimes web search. That is why an AI assistant and the model running it often have different names.
The EU AI Act reflects this distinction too: it treats “general-purpose AI models” separately from the “AI systems” that build on them.
Families of models
- Large language models (LLMs) produce text and code.
- Image generation models create pictures from a description.
- Speech recognition models transcribe what people say.
- Multimodal models combine several of these abilities.
When a trained model answers your requests, that step is called inference. It costs money every time, which explains the usage caps on free plans.
Why it matters to you
Two apps with different names may run on similar models, and the same app can give you very different answers depending on the model picked in a menu. When you compare tools, look at the model as well. The model comparison tool lets you line up prices, capabilities and context size side by side.
Common misconceptions
“ChatGPT is a model.” It is an app; it runs on models from OpenAI’s GPT family.
“The model learns from every chat in real time.” A released model is frozen. Improvements arrive with a new version.
“Open models are always worse.” Some open-weight models hold their own against closed ones on many tasks; it all depends on what you need.
Related terms
- Large language model (LLM) · A program trained on huge amounts of text to predict what comes next in a text; it is the engine behind AI assistants.
- Model parameters · The internal numbers a model tunes during training; how many it has indicates its size, not directly how good it is.
- AI assistant · A consumer app such as ChatGPT, Claude or Gemini that answers, writes, summarizes and helps with everyday tasks using AI.
- Training data · The examples (text, images, audio and more) an AI model learns from; they shape what the model knows and where it goes wrong.
- Open-weights model · An AI model whose trained parameters (the weights) can be downloaded, so you can run it on your own hardware.
- Inference · The moment a trained model is actually used to produce an output: every question you send to an AI triggers an inference run.
Try it
Full glossary
Frequently asked questions
What's the difference between an AI model and an AI assistant?
The model is the trained computing engine. The assistant is the app that uses it and adds an interface, memory, web search or file uploads. One assistant can switch models, and one model can power several apps.
Why do model names change so often?
Labs regularly train new versions that are faster, cheaper or more capable, and give them a new number. The app often keeps its name while the model under the hood changes.
Can I run an AI model on my own computer?
Yes, for some so-called open-weight models whose files are published. You need a fairly powerful machine, especially in graphics memory. Most of the leading models from big labs are still only available online.
Is an AI model a database?
No. It does not hold a list of ready-made answers. It has learned patterns and computes a fresh answer for each request, which is why it can write a sentence nobody wrote before, and also why it can be wrong.
Data sources
Data as of October 22, 2026
Report an error
Does a figure, a date or a fact look wrong? Tell us: every report is checked.