How much text a model can see at once — your whole conversation, pasted documents, and its answer. OMA offers models with up to 1M token windows: enough to read an entire book in one go.