Reliable software tools for mathematical modeling have been developed for simulation models but for constrained optimization models progress has been slower. The following approach examines the possibility of working with large-scale mathematical models in a compact form permitting the direct analysis of input-output relations where information is transformed into a compressed mode (as opposed to aggregation procedures) without loss of accuracy. The article also describes the approach when applied to discrete dynamic models.