Model limits
Context window vs. output limit
Model documentation often lists a large context window beside a smaller maximum output. Those numbers answer different questions, and neither means you can safely fill the entire window with a new prompt.
The context window
The context window is the model's working space for a request. Depending on the provider and API, it can include system instructions, user input, earlier turns, tool calls, retrieved documents, reasoning tokens, and generated output.
The output limit
The output limit caps how many tokens the model can generate in one answer. A model with a 1M context window and a 128K output cap cannot produce a 1M-token answer. It may also stop earlier because of your request settings or the provider's product limits.
Plan with a reserve
Choose a response reserve based on the task: a short extraction needs less room than a long report or a large code patch. Then subtract existing conversation usage. What remains is the practical budget for your new prompt and files.
- Short classification or extraction: reserve roughly 1K–4K tokens
- Detailed analysis or writing: reserve roughly 8K–16K tokens
- Large structured output or code generation: reserve more and validate the provider cap