Context window
The amount of text a model can hold in mind at once, counting both what you give it and what it writes back. It is measured in tokens. Once a conversation or document runs past the limit, the model starts losing track of the earliest parts. This explains a common frustration, that a tool seems to forget something said earlier, and it has practical consequences in healthcare: a long record or a lengthy guideline may not fit in one window, so what the model can reason over at any moment is bounded. Larger context windows cost more to run, which is part of why capability and price move together.
A pre-clinic summary tool works well in its pilot clinic, where most patients are new referrals with short records. In oncology, for patients with years of notes, its summaries cover recent scans in detail and omit the original staging and an infusion reaction from the first cycle, because the oldest documents no longer fit in what the model is given. The summary reads just as fluently, so nothing warns the reader. Ask how the tool selects or condenses documents when a record exceeds its window, and test it on your longest records.
Terms like this come up in real clinical scenarios across the HelloAI courses: bite-sized modules with verifiable certificates. An account takes one minute, no password needed.
Sign in →