The development of novel polymeric membrane materials for high performance gas separation applications is discussed in this dissertation. Membrane-mediated gas separations have been demonstrated to be a more energy efficient method compared to conventional separation techniques such as distillation. In the past few decades, numerous research efforts have been focused on new polymer designs to overcome the inherent tradeoff between permeability and selectivity by constructing contorted backbone or introducing bulky units to create low-barrier pathways for fast gas transport. While the permeability-selectivity tradeoff seems to be tackled by many new high-performing polymers pushing the upper bound limits further toward the upper right side, polymer gas separation membranes frequently suffer from physical aging (e.g., loss of permeability over time) and plasticization (e.g., loss of size sieving in the presence of condensable gases like CO2). The impact of these challenges is clearly evidenced by the fact that only a handful of polymers have made it into commercial applications despite hundreds of new polymers have been developed and evaluated for gas separation membranes. At molecular level, intolerance to temporal changes of existing size-sieving polymers largely originate from the transient nature of non-equilibrium conformational free volume that is susceptible to external changes due to chain relaxation or increased chain mobility. As such, the search for the next-generation polymer membrane materials should be directed toward constructing well-controlled and stable polymer free volume architecture, where segmental motion or chain relaxation is restricted or suppressed.Crosslinking has been well-acknowledged as an effective way to construct stable free volume. However, the existing crosslinking has problems of reduced permeability, unpredictable performance, and very limited tunability as a result of crosslinking-induced densification and complicated network structures. To this end, we propose a new approach of end-linking in combination with phenylethynyl crosslinking chemistry to address these issues via constructing crosslinked model network structure where crosslink density and inhomogeneity can be feasibly tuned. In contrast to conventional random crosslinking approaches, three distinct crosslinked model networks (i.e., unimodal, bimodal, semi-interpenetrating networks) are prepared via thermal crosslinking of specifically designed phenylethynyl-terminated telechelic oligomers with systematically varied molecular weights, where the comprehensive study about two independent parameters, crosslink density and crosslink inhomogeneity, were exemplified in this work.In this dissertation, we first tested this hypothesis in commercial Matrimid®-like polyimide with unimodal network (Chapter 2), where markedly improved permeability and well-maintained selectivity was observed in this unimodal system compared to their linear counterparts. This unique design of crosslinked model network was further applied to novel iptycene-based polymers, including crosslinked pentiptycene-containing polybenzoxazole (Chapters 3, 4 and 5) and triptycene-containing polyimide (Chapter 6). Besides the interplay between the competing effect of crosslinking-induced densification and bulky crosslinkable end-groups, it is clearly demonstrated for the first time that crosslink inhomogeneity contributes to gas transport apart from crosslink density by disrupting segmental packing as evidenced by the greatly improved permeability and enlarged FFV of bimodal networks. Lastly, we extended the model network concept to construct semi-interpenetrating (sIPN) structures in PBI/Matrimid system (Chapter 7). The sIPN approach expands the materials spectrum, while the PBI/Matrimid demonstrate promising separation performance at high temperatures.