AI EncyclopediaCore concepts
Large Language Model (LLM)
/large LANG-gwij MOD-el/
A deep-learning model, usually built on the transformer architecture, trained on huge amounts of text to predict the next token and, in doing so, learn grammar, facts, reasoning patterns and style.
In plain English
An LLM is software that has read an enormous amount of text and learned to predict what word (or word-fragment) comes next. That simple skill, done at massive scale, turns out to be enough to write, summarise, translate, answer questions and hold a conversation - without anyone hand-coding rules for any of it.
Technical explanation
An LLM is a neural network, almost always transformer-based, trained via self-supervised learning to estimate the probability distribution of the next token given prior context across a huge, mostly unlabelled text corpus. Training happens in stages: large-scale pre-training on raw text, then post-training (instruction tuning, RLHF, or both) to align outputs with what a user actually wants. Models are characterised by parameter count - typically billions to trillions - and, since 2023, increasingly by mixture-of-experts routing rather than pure dense parameter scaling.
Why it matters
LLMs are the substrate almost every current AI product sits on top of - chat assistants, coding tools, search summarisation, enterprise copilots. Understanding what one is (a next-token predictor, not a database of facts) is the single most useful piece of AI literacy for any organisation deciding where to trust one and where not to.
Real-world example
GPT-4o (OpenAI), Claude (Anthropic), Gemini (Google) and Llama (Meta) are all LLMs; Llama 2, released July 2023, shipped with under half the parameters of GPT-3 while its backers claimed comparable accuracy - illustrating that parameter count alone does not determine capability.
Common misunderstanding
That an LLM 'looks things up' the way a search engine does. It does not query a live knowledge base by default - it generates text token by token from patterns learned during training, which is exactly why it can state a confident, fluent, and wrong answer (see Hallucination).
Something wrong here?
Every entry is hand-researched and hand-written by Nakoda. If a fact is stale, a source has changed or a definition needs sharpening, tell us and we will check it.