← Back to Glossary
Context Window
Definition
A context window is the maximum amount of information, measured in tokens, that a language model can consider at once during a single inference step.
Purpose
The purpose of the context window is to bound the information available to the model at inference time, directly influencing what the model can attend to, reason over, and respond based on.
Key Characteristics
- Fixed or model-dependent maximum token capacity
- Includes system instructions, user input, retrieved data, and intermediate context
- Limits the amount of information a model can directly reason over in a single inference
- Requires prioritization, summarization, or pruning of context in complex systems
- Distinct from long-term storage or agent memory mechanisms
Usage in Practice
In practice, the context window is managed by selecting, ordering, and compressing inputs so that the most relevant information fits within the model's token limits during each inference call.
One implementation of this concept is offered by Kenaz through the Semantic Engineering service.