Context window

A context window is the maximum number of tokens a large language model can process in a single call — including both the input prompt and the generated output.

Modern LLMs have context windows ranging from 8,000 tokens (older or smaller models) to 1 million tokens (frontier models with long-context support). Larger windows let the model "see" more brand context at once: longer system prompts, more past posts retrieved by RAG, longer in-progress documents.

Context window is a hard ceiling. A tool that wants to inject 100 past posts into the prompt for retrieval-augmented generation cannot do so against a model with an 8K window without summarising heavily. Tools that handle long-context well typically either select context aggressively or rely on retrieval to pick only the most relevant pieces.

Brand-conditioning quality is bottlenecked by how much brand-specific context the tool can fit into the model’s context window. Tools that don’t respect the window blow past it and silently truncate; tools that engineer for it produce more consistent output.