Large Language Model (LLM)

A large language model is a type of AI system trained on massive amounts of text that can understand, summarize, translate, and generate human language with remarkable fluency.

The name gives you the shape of the thing. Large refers to scale — both the volume of text these models are trained on and the number of internal settings, called parameters, the model uses to make sense of it all. Language refers to what it works with — words, sentences, paragraphs, conversations, code. Model refers to the mathematical system that has learned the patterns and structure of human language by processing billions of words drawn from books, websites, code repositories, and more. At the heart of every modern large language model is an architecture called the transformer, which allows the model to weigh relationships between words across an entire passage of text rather than processing them one at a time.

Training a large language model works by having the system process enormous quantities of text and repeatedly practice predicting the next piece of text — known as a token — in a sequence. Over billions of these predictions, and billions of corrections when it gets them wrong, the model builds up a deep internal representation of how language works. By the time training is complete, it has developed something that functions like a broad, flexible understanding of the world as expressed through text. It does not think or reason the way a human does, but it has seen so many examples of human reasoning written down that it can produce outputs that look remarkably similar.

What makes large language models significant is their versatility. The same model that can turn a rough idea into a polished essay can also write Python code, translate between languages, answer questions, draft emails, and hold a coherent conversation — without being built specifically for any of those tasks. This general-purpose capability is what separates them from earlier, narrower AI systems that were built to do one thing well. ChatGPT, Claude, and Gemini are all built on large language models, which is why they feel so different from the chatbots and voice assistants that came before them.