What's an LLM? The model behind ChatGPT, Claude and Gemini

An LLM (Large Language Model) is an artificial intelligence model trained on huge amounts of text, able to understand and generate language ^{1, 2}.

OpenAI's GPT, Anthropic's Claude and Google's Gemini are families of LLMs. ChatGPT, the Claude app and the Gemini app are the products built on top of them.

How it works, without formulas

During training, the model reads more text than any person could read in a lifetime, and learns how words tend to follow one another. When you write to it, it produces the answer one piece at a time, each time choosing the most plausible continuation.

This is where its strength comes from (it writes, summarises, translates, reasons about a text) and also its limit: plausible does not mean true. An LLM that has no access to your price lists can make up a price with the same confidence with which it writes an email.

Why it matters to you

Many AI tools for businesses have no model of their own: they use an LLM from one of the big producers and build an interface on top of it. This is not a flaw, it is the normal way of working. But it means two things.

The first: when you compare two suppliers, ask which LLM they use. Two products with very different prices can have the same engine.

The second: the quality of a tool depends less on the model and more on how it is connected to your data. An LLM that reads your manuals and your catalogue answers better than a more powerful one that has never seen them.