What Is a Context Window?

A model's context window is the maximum number of tokens it can consider in a single request — combining everything you send in and everything it sends back. If your total exceeds that limit, the request is typically truncated or rejected.

What counts toward the context window

• The system prompt (instructions, persona, constraints).
• The current user message.
• Prior conversation turns, if you're sending history.
• Retrieved context, in RAG applications.
• Tool or function definitions and their outputs.
• The model's own output tokens.

Is high usage always a problem?

Not necessarily. There's no universal threshold at which usage becomes unsafe — it depends on the model and what your application still needs room for. A single-turn summarization task can comfortably run near the limit; a multi-turn chatbot needs headroom for the conversation to keep growing.

Use the Context Window Calculator to see how your specific prompt structure adds up against a given model's limit.