A realistic and computationally efficient method to model fluid adsorption in disordered mesoporous glasses is described. Local mean-field theory of a lattice model of adsorption is combined with the representations of the disordered matrix configurations generated by Gaussian random fields. The experimental structure factor and porosity of the glass serve as input. The method allows for an efficient route to perform disorder averages for large systems. The adsorption and hysteresis behaviors obtained agree closely with those observed experimentally.