Large Language Model — a program that learned patterns from massive amounts of text and uses them to produce new text, one word (token) at a time.
↑ ContentsDefinition
An LLM receives text as input (the context) and produces text as output (a response). It does not execute code, read files, or access the internet on its own. In agent systems, the LLM is the model component: the runtime feeds it context and interprets its responses, turning token prediction into action. LLMs generate markdown and JSON well because most of their training data — GitHub READMEs, documentation, Stack Overflow — is structured in these formats.
↑ Contents