Model order reduction (MOR) is a promising research area as MOR techniques help in simplifying complex physical systems. This paper discusses prominent classical MOR methods and evolution of soft-computing techniques applied to the large-scale systems for reducing their order. The state-of-the-art discussion highlights that optimization methods provide more efficient reduced-order models. To gain insights, a 6th order proton exchange membrane fuel cell (PEMFC) system and its 2nd order reduced model are discussed using existing classical and optimization techniques. The comparison reveals that the soft-computing techniques such as Grey Wolf Optimization and Particle Swarm Optimization offer least integral square error of 10−7 for a 2nd order model. In addition, time-domain parameters such as rise -time, settling-time and other parameters of reduced system are approximately closer with those of higher-order system. This demonstrates that soft-computing techniques not only enhance accuracy but also maintain the essential dynamic characteristics of the higher-order system which include stability and minimum phase nature. Future work could explore the application of these methods to even higher-order systems and other domains, setting their efficacy in complex system modeling.