Large Language Model (LLM)
A large language model (LLM) is an AI model trained on very large amounts of text to predict and generate language, which lets it answer questions, summarise, translate and hold a conversation.
Definition
A large language model is a neural network trained on huge collections of text to learn the patterns of language. Given some text, it predicts what should come next, one piece at a time, and by doing this it can write, summarise, translate, classify and answer questions. Well-known assistants are built on such models.
Their output depends on the instructions they receive, called the prompt, and on the information supplied with it. They are evaluated on tasks relevant to your use: accuracy of answers, consistency across repeated runs, handling of your languages, and how they behave on unusual or sensitive requests. Many applications give the model reference documents to rely on, so that answers come from your content.
The main limitation is that an LLM generates plausible text, not verified facts, so it can state something false with confidence. It can also reflect biases found in its training data and cannot explain its reasoning with certainty. For decisions affecting people, use it to draft, structure or summarise, and keep a person responsible for the outcome.
Key points
- It is trained on large amounts of text to predict and generate language.
- It can summarise, translate, classify and converse, depending on how it is instructed.
- It can produce fluent but incorrect statements, sometimes called hallucinations.
- It can reflect biases from its training data, so test it on your own cases.
- Keep a person responsible for any decision that affects someone.
Example
A customer success team uses an LLM to turn a 15-minute recorded customer conversation into a five-line summary. A manager compares a few summaries with the recordings every month, and notices that the model sometimes merges two separate complaints, so the team adds a rule to list them separately.
Frequently asked questions
Is an LLM the same as artificial intelligence?
No. An LLM is one type of AI model, specialised in language. AI also covers image recognition, forecasting and many other techniques.
Does an LLM understand what it says?
It handles language very well, but it does so by statistical patterns rather than by checking facts. That is why answers should be verified when they matter.
Can an LLM be used on personal data?
It can, provided the data protection rules that apply to you are respected: a lawful basis, limited retention and clear information to people. Check the provider's terms and ask a legal advisor.
Related terms
AI conversations, human decisions
Passisto holds structured voice conversations with candidates, employees, clients and students, and turns them into reports your team can act on.