Artificial Intelligence & AI Agents

What Is a Large Language Model (LLM)?

Definition

A large language model (LLM) is an AI model trained on very large amounts of text to understand and generate human language. It predicts the most likely next piece of text given what came before, which lets it answer questions, summarise documents, translate, write code and follow instructions. Well-known examples include the models behind ChatGPT, Claude and Gemini.

Jump to FAQs

Also known as LLMs, foundation model, language model, GPT model

What does an LLM mean in simple terms?

Imagine an extremely well-read autocomplete. An LLM has seen so much text that, given a question or instruction, it can predict a sensible, well-written continuation. Because language carries knowledge and reasoning, that prediction is often a useful answer.

The word large refers to the size of the model, often billions of internal parameters, and to the scale of the data it learned from. That scale is what lets one model handle many different tasks without being retrained for each one.

How do LLMs work?

LLMs are built on a neural network design called the transformer. In practice:

  1. 1

    Tokenisation

    Text is split into tokens, small chunks of words. Pricing and limits for most LLMs are measured in tokens.

  2. 2

    Pre-training

    The model reads a huge corpus and repeatedly learns to predict the next token, building a statistical picture of language and facts.

  3. 3

    Instruction tuning

    Further training on curated examples and human feedback teaches it to follow instructions and respond helpfully and safely.

  4. 4

    Inference

    At run time, the model receives a prompt within its context window, the maximum amount of text it can consider at once, and generates a response token by token.

  5. 5

    Integration

    Applications call the LLM through an API, often adding company documents via retrieval-augmented generation or tools that let it take actions.

What are real-world examples of LLMs?

Contract review

A legal team asks an LLM to extract renewal dates, liability caps and termination clauses from supplier contracts.

Support triage

Incoming tickets are classified by topic and urgency, and a draft reply is prepared for the agent to approve.

Data questions

A manager asks, in plain English, which regions missed target last quarter, and the LLM writes and runs the database query.

Code assistance

Engineers use an LLM to explain unfamiliar code, generate tests and suggest fixes.

Why do LLMs matter for businesses?

LLMs are the engine behind most current generative AI and AI agent products. Choosing a model, and deciding whether to use a hosted API or a privately deployed model, affects cost, speed, quality and data privacy.

For technology leaders, the key decisions are which tasks suit an LLM, how to ground it in trusted data, and how to evaluate its answers before they reach customers.

What are the benefits and challenges of LLMs?

Key benefits

  • One model handles many language tasks without separate training
  • Understands natural-language instructions, so non-technical users can work with it
  • Strong at summarising, drafting, extracting and translating
  • Can be combined with tools and data to power agents and assistants
  • Available as hosted APIs or as open-weight models that can run privately

Challenges and limitations

  • Can hallucinate facts, so answers need grounding and checks
  • Knowledge is fixed at training time unless current data is supplied
  • Context windows limit how much information fits in one request
  • Token costs and response times matter at scale
  • Sensitive data needs controls, or a private deployment, to meet compliance needs

How is an LLM different from related terms?

LLM vs Generative AI

Generative AI is the category of AI that creates content. An LLM is the specific kind of model that generates text within that category.

LLM vs NLP

NLP is the whole field of computers working with language. LLMs are the most powerful recent technique within NLP.

Frequently asked questions about an LLM

What does LLM stand for?

LLM stands for large language model, an AI model trained on large amounts of text to understand and generate language.

Is ChatGPT an LLM?

ChatGPT is an application. It runs on large language models developed by OpenAI. The LLM is the underlying model; ChatGPT is the product people use to talk to it.

What is a context window in an LLM?

The context window is the maximum amount of text, measured in tokens, that a model can consider in one request, including both the prompt and its response.

Can a company run its own LLM privately?

Yes. Open-weight models can be deployed on a company's own cloud or servers, so prompts and data never leave its environment. This usually trades some capability and adds infrastructure cost for greater control.

How do you make an LLM answer from company data?

The usual approach is retrieval-augmented generation: relevant passages from company documents are retrieved and given to the model with the question, so it answers from those sources.

How Stack18 can help

Stack18 integrates hosted and private LLMs into business systems, with evaluation and guardrails before anything reaches users. See OpenAI and Azure OpenAI integration, private LLM deployment and LLM evaluation and quality assurance.