In this paper, an optimization-based model orderreduction (MOR) framework is proposed. The method involvessetting up a quasi-convex program that solves a relaxation of theoptimal H∞ norm MOR problem. The method can generate guaranteedstable and passive reduced models and is very flexible inimposing additional constraints such as exact matching of specificfrequency response samples. The proposed optimization-basedapproach is also extended to solve the parameterized modelreductionproblem (PMOR). The proposed method is compared to existing moment matching and optimization-basedMOR methodsin several examples. PMOR models for large RF inductors oversubstrate and power-distribution grid are also constructed.