A mixture of aromatic hydrocarbons, comprising ethylbenzene and at least one xylene, is isomerized using a two component catalyst system to convert the ethylbenzene to compounds that may be removed from the aromatic hydrocarbon stream and to produce a product stream wherein the para-xylene concentration is approximately equal to the equilibrium ratio of the para-isomer. The first catalyst comprises an intermediate pore size zeolite that is effective for ethylbenzene conversion. The first catalyst is preferably silica-bound. The second catalyst comprises an intermediate pore size zeolite, which further has a small crystal size and which is effective to catalyze xylene isomerization. Each of the catalysts of this invention may contain one or more hydrogenation/dehydrogenation component.