Long context

The capability to accept unusually large request contexts, with the practical size determined by the model, modality, product, and output budget.

Long context can allow a large book, code selection, or document collection to enter one request when it fits the token budget. Performance varies with task, position, redundancy, conflicts, and model. Lost-in-the-middle and long-context evaluation research shows that capacity and retrieval quality are separate properties, while newer models can improve on earlier patterns. Test the current configuration rather than treating either failure or success as permanent.

Builder example

Long context, retrieval, summarization, and hierarchical processing are alternative or complementary designs. One may outperform another depending on how much global structure the task needs, how often sources change, latency, cost, and which claims must be cited. Compare complete designs on representative long inputs.

You paste a two-hour meeting transcript into a chat and ask what was decided about the budget. The decision was in the middle, and the model misses it.

Search for the relevant section first, then ask about that section specifically. Targeted questions get better answers than broad ones over large inputs.

Common confusion: Long context concerns one request's capacity. Persistent product memory, project files, and conversation state are separate systems that may load selected information into later requests.