What is an LLM?

Think of an AI model as a tool trained to find patterns and produce an answer. Different models are good at different jobs. The one behind a text conversation is often an LLM: a large language model. That name tells you its main job is language.

What does an LLM do?

It works with words: your question, the text you attach, and the reply it writes. Most chat LLMs build a response a small piece at a time based on the text so far. That lets them explain an idea, draft an email, or summarize notes. “Large” refers to the scale of the trained model, but there is no single size at which a language model officially becomes an LLM.

The chat app is not the same thing as the model. An app can put search, file access, voice, or other features around one or more models. That is why two apps using similar models can feel different.

Are there other kinds of models?

Yes. An image model might identify what is in a photo or create a picture. A speech model might turn spoken words into text or speak text aloud. A video model might analyze a clip or generate one. A prediction model might sort a message as spam or estimate a number. These are examples of jobs, not a perfect set of mutually exclusive categories.

What does “modality” mean?

It just means a kind of information. Text is one modality. Pictures, sound, video, and numbers are others. If you type a question and attach a photo, you have supplied two modalities: text and image.

A multimodal model works with more than one kind. But check the direction: a model that can read photos may not be able to make photos. “Multimodal” alone does not tell you its exact inputs and outputs.

Where does an agent fit?

An AI agent is not another model type in this list. It is a way of using a model with tools and several steps to do a job. A language model might be the part that understands the request, while the surrounding system searches, checks results, and asks for your decision.