Context window
The context window limits how many tokens a language model can consider in one processing step. It is not a permanent knowledge archive.
Distinguish available context from stored knowledge
The context window describes the limited number of tokens a language model can consider in one processing step. This may include instructions, earlier messages and supplied document excerpts. The application also determines which content actually enters it.
In a fictional example, an assistant must compare a current delivery rule with an older version. If only the old version is supplied, a larger context window alone does not help. It increases possible input capacity but does not obtain the missing current source.
The context window is not a permanent memory of every company document either. Search or a document connection can supply relevant content for a request. Even if a long text fits, it remains necessary to check whether the decisive information was used correctly in the result.
Sources
Related terms
- Language model tokenTokens are processing units for language models. They do not reliably correspond to whole words and differ from design tokens.
- Large language modelA large language model processes language using learned patterns. Fluent text proves neither factual accuracy nor access to current business data.
- PromptA prompt is the input to a generative model. A clear task helps but does not replace data connections or access permissions.
- Model cardA model card documents an AI model’s purpose, foundations, evaluation conditions and known limitations for a particular version.
- AI readinessAI readiness concerns the foundations of an AI project: objectives, usable data, skills, responsibilities and an operating workflow.
- Context engineeringContext engineering prepares information for individual model calls: relevant instructions, messages, sources and current states.
Articles on this topic
AI & Automation with ASES
AI assistants and automated workflows