I would expect a flash model, with its reduced size, to suffer on tail tasks. That is the trade-off you make.