Send us the problem Free AI check
Currently being edited
in Johannesburg
The morning AI brief,
every day

Learn AI4 min read

LLM meaning: what a large language model is, in plain English

Illustration: In a tall library a man with brown skin on a rolling ladder pulls one book from a wall of books that rises out of the frame.

The LLM meaning in artificial intelligence is large language model: the technology behind ChatGPT, Claude, Gemini and most AI assistants. In South Africa the letters also mean a Master of Laws degree, which is why the question comes up so often. Here is what an LLM is, how it works, what it does well and badly, and what it means for a business.

A large language model is software trained on enormous amounts of text to predict the next word, and that simple skill turns out to be enough to write, summarise, translate, answer questions and write code. It does not look facts up unless it is connected to search or to your documents, which is why it can sound confident and still be wrong.

For a business, the useful question is not “which LLM” but “what job”: answering customers, reading documents, drafting quotes. The model is a component; the value is in connecting it to your data and your process. ZAIQ, the company behind AI Guide, does that work: see AI engineering.

What LLM stands for, in AI and in law

In AI, LLM stands for large language model. “Large” refers to the size of the model, billions of internal settings learned during training, and to the size of the text it learned from. “Language model” means its core skill is working with language.

In South African universities, LLM also stands for Legum Magister, the Master of Laws postgraduate degree offered by universities such as UNISA and Wits. If you arrived here looking for the law degree, that is the other LLM; the rest of this page is about AI.

How a large language model works

The short version, without the maths.

  1. Training on text. The model reads a huge collection of text and learns patterns: which words and ideas tend to follow which.
  2. Predicting the next word. Given some text, it predicts what comes next, one small piece (a token) at a time. Long answers are built from many of these predictions.
  3. Tuning for helpfulness. After training, it is tuned with human feedback so it follows instructions and answers helpfully and safely.
  4. Connecting to tools. Modern assistants connect the model to search, documents and apps, so answers can use current facts instead of memory alone.

What LLMs are good and bad at

Good atWatch out for
WritingDrafting emails, quotes, summaries and reports quickly.Confident errors if nobody checks the facts.
ReadingPulling key terms, dates and figures from long documents.Missing context that is not in the text it was given.
AnsweringAnswering common customer questions instantly, any hour.Inventing answers when it has no source: always ground it in your data.
CodingWriting and fixing code, fast.Subtle bugs that need a developer to review.
Maths and factsExplaining concepts and reasoning step by step.Exact figures, unless it uses a calculator or a trusted source.

Examples of LLMs in November 2026

These are the large language models most South Africans meet, through the apps built on them. Model names change every few months.

  • OpenAI GPT-6. GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna, the models behind ChatGPT. The free plan runs on Luna. (ChatGPT)
  • Anthropic Claude. Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5. The free plan uses Sonnet and Haiku. (Claude)
  • Google Gemini. Gemini 4 Argon on paid plans, with lighter Flash-Lite on the free plan. (Gemini)
  • DeepSeek. DeepSeek-V4.1-Flash and V4 Pro, from a Chinese company, with open weights for some models. (DeepSeek)
  • Meta. The models behind Meta AI, the assistant inside WhatsApp, Facebook and Instagram. (Meta AI)

LLM terms you will hear

  • Prompt. The instruction or question you give the model. (Basics)
  • Token. A small chunk of text, roughly three quarters of a word, that models read and write and that APIs charge for. (Basics)
  • Context window. How much text the model can consider at once, from a page to whole books. (Capacity)
  • RAG. Retrieval-augmented generation: the model looks things up in your documents before answering, which cuts made-up answers. (Business)
  • Agent (agentic AI). An AI that takes actions, such as sending messages, updating systems or running steps on its own, not just answering. (Business)
  • Hallucination. A confident answer that is wrong or invented. (Risk)

What it means for your business You do not need to build an LLM. You need to aim one at a real job and connect it to your data safely, with POPIA in mind and a person checking anything that matters. See ChatGPT versus a custom AI agent and whether AI is worth it for a South African business.

Questions people ask

What does LLM mean in AI?

Large language model: an AI trained on large amounts of text to understand and generate language. ChatGPT, Claude and Gemini are built on LLMs.

What does LLM stand for?

In AI, large language model. In law, Legum Magister, Latin for Master of Laws. The context tells you which one is meant: a chatbot article means the AI model, and a university prospectus means the degree.

What is the difference between AI and an LLM?

AI is the whole field of getting computers to do things that seem intelligent. An LLM is one kind of AI system, trained on text. So every LLM is AI, but most AI, such as fraud detection or image recognition, is not an LLM.

What does LLM mean in law?

Master of Laws (Legum Magister), a postgraduate law degree offered by South African universities including UNISA and Wits.

Is ChatGPT an LLM?

ChatGPT is an assistant built on OpenAI's large language models, with extra tools such as search, file handling and image generation added around the model.

Can a business use an LLM safely?

Yes, if it is connected to the right data, kept inside controlled accounts, and checked by a person for anything involving money, law or health. POPIA applies to any personal information it handles.

Sources

  1. What are large language models (LLMs)? (IBM)
  2. Attention Is All You Need (Vaswani and others, arXiv, June 2017)
  3. OpenAI Help Center: Model release notes
  4. Models overview (Claude Docs, Anthropic)
  5. Gemini models (Google AI for Developers)

Checked November 2026. Prices and features change, so confirm on the official site before you buy.