Context window
The maximum amount of text, measured in tokens, that a model can take as input and attend to when producing a single response.
In more detail
The context window is the model's working memory for a single call. Everything the model considers, including the instructions, the retrieved source material, and the conversation so far, has to fit inside it. Larger windows let a step reason over more material at once, such as a long contract or a batch of records.
A large window is not a substitute for good retrieval. Stuffing everything into context is slower, costlier, and can dilute the model's attention. The reliable pattern is to retrieve the relevant material and pass that into the window, rather than pasting in everything and hoping the model finds what matters.
Where this shows up at Ceven
Ceven's research and document steps select the relevant material to place in a model's context window rather than dumping raw source in wholesale. Wide and deep research gather and rank sources first, then reason over the ones that matter, which is how the resulting briefs stay grounded and cite where each claim came from.