What is an LLM?

A large language model (LLM) is an AI model trained on vast amounts of text to predict the next token, which lets it generate and transform language: writing, summarising, classifying, answering questions, and calling tools. It is the engine behind most modern AI features.

Read in RomanianLLM (model lingvistic mare)

Why it matters

The LLM is the component most people mean when they say "AI" today. Picking the right one, and wiring it up well, decides whether an AI feature feels sharp or flaky. The model is one part of the system, though; retrieval, prompts, and guardrails usually matter more than which model you choose.

What it is good and bad at

LLMs are strong at language tasks: drafting, rewriting, extracting structure from messy text, and routing. They're weak at being a reliable source of facts on their own, which is why production systems pair them with RAG, and at staying in a fixed format, which is where fine-tuning helps.

Where teams get it wrong

Treating the LLM as a database. It predicts plausible text, not verified truth, so without grounding it will state wrong things fluently. Design for that from day one rather than discovering it in front of a customer.

Related terms

← Back to glossary