Context window

The most text a model can consider at once, counted in tokens. Anything longer has to be cut, summarised, or retrieved, which is why a long document is not simply pasted in.

• See also: Token
• See also: Retrieval augmented generation
• Used in: LLMs

Back to the glossary