Reverse osmosis (RO) is a key technology to various separation applications, whose processes traditionally rely on the well-known size-exclusion mechanism to block the greater component in size. Interestingly, recent studies have also demonstrated RO processes that can selectively permeate larger molecules via their favorable adsorption. While these distinct behaviors have been observed, a comprehensive understanding on the design of RO membranes remains lacking. To this end, by employing molecular dynamics simulations, this study conducts a systematic evaluation on RO membranes with different pore dimensions for their separation performance of a binary mixture comprising components of varying molecular sizes, offering quantitative insights into the design of RO membranes. The outcomes show that, while the traditional size-exclusion design remains an effective approach, utilizing the adsorption-facilitated mechanism may lead to a notably enhanced performance with rich separation behaviors. Specifically, the pore size may be designed to be greater than both of the components for a boosted permeation flux. Moreover, through strategically exploiting the confinement effect for preferentially adsorbing a specific component, either the smaller or the greater one, effective RO processes to selectively permeate the either component can be achieved.