Large Language Model
LLMAlso called language model.
A very large model trained on huge amounts of text to predict what comes next, which lets it chat, summarise, translate, and draft.
Why it matters
It's the engine behind the assistants most people now use daily. Understanding that it predicts plausible continuations — rather than looking up facts — explains both its fluency and its mistakes.
An example
A consultancy drafts meeting recaps from raw notes with an LLM, then a person checks the commitments and dates before the summary is sent to the client.
Where you encounter LLMs
An LLM is an underlying model; an assistant is the app you interact with. An assistant may use different models, tools, and search systems depending on the task and product setting.
Related glossary ideas: , , and .
ChatGPT
By OpenAI
A conversational assistant for writing, analysis, research, coding, and working with files and images.
Claude
By Anthropic
An assistant for conversation, writing, analysis, coding, and working through documents and other material.
Gemini
By Google
Google’s assistant for asking questions, creating content, researching topics, and working across multiple media.
Microsoft Copilot
By Microsoft
Microsoft’s assistant experience for search, drafting, creating, and help across Microsoft products.
More examples
Meta AI
By Meta
Meta’s assistant, available on the web and within several Meta apps, for questions, ideas, and creation.
Grok
By xAI
An assistant for conversation, search, research, writing, coding, and creating media.
Vibe (formerly Le Chat)
By Mistral AI
Mistral AI’s assistant for conversation, research, drafting, and working with documents.
Perplexity
By Perplexity
An answer and research assistant that searches sources and can offer access to different underlying models.
Illustrative selection, not a ranking or recommendation. Product marks belong to their respective owners and are shown for identification.
How this connects
Helpful to know first
Related learning
Go deeper
Training exposes the model to enormous text corpora and adjusts billions of parameters so it assigns high probability to likely continuations. Later stages tune it to follow instructions and to behave within policy, but the underlying operation stays next-token prediction.
Sources
Was this clear?
Explore related terms
Select a term to see its definition.
Helpful to know first
Read these before this one.
Related ideas
Concepts used alongside this one.
Explore next
Next in AI essentials.