What is a large language model?
A large language model (LLM) is an AI system trained on enormous quantities of text to predict what words should come next. By learning those patterns across billions of passages, an LLM becomes able to understand prompts, answer questions, summarize, translate, and generate coherent writing and code.
LLMs are built on deep learning — specifically a neural-network architecture called the transformer — and they're the technology behind modern chatbots and writing assistants.
How they work
At their core, LLMs are very sophisticated next-token predictors: given a sequence of text, they estimate the most likely next piece. At scale, that simple objective produces surprisingly capable behavior — following instructions, reasoning across steps, and adapting tone.
What they're used for
LLMs power chatbots, coding assistants, search experiences, document summarization, translation, and the writing features built into many products. They're also combined with other tools to take actions, not just produce text.
Strengths and limits
LLMs are remarkably flexible and useful for drafting, explaining, and brainstorming. But they can confidently produce wrong information (hallucination), reflect biases from training data, and lose track over long contexts. The best applications pair an LLM with grounding — facts, documents, or tools — so outputs are checked rather than trusted blindly.