Agreed. That said, in general a 1M context model has a larger usable window than a 260k context model.