There are a large number (>125) of molecular representations for coals that span the rank range over seven decades. However, their utility has mostly been in representing chemical structural features, rather than in probing physical structure or exploring the structure-behavior relationship. This paper examines the utility of coal models and reviews the existing and emerging opportunities for coal models to contribute to coals effective utilization via demystification of the structure-behavior relationship. Coal models have been used to explore the coalification pathway, including contraction with water removal. Physical evaluations have probed the density of models as a check on their accuracy. Pore size distribution and sorption have been explored in simple pores and more recent work with carbon dioxide, water and methane sorption within the porous structure of large-scale (<20,000 atoms) model. Pair distribution frequency and small angle X-ray scattering simulations have also been compared with experimental observations and offer an additional check on the constitution of the model structure. Simulated HRTEM and simulated (calculated) NMR spectra also exist. Models have been disassembled in efforts to represent the pyrolysis process, char formation, and char reactivity (including the role of ion-exchangable ions). Similar to the pyrolysis models, direct liquefaction has been explored with a pyrolysis style approach. Coal-solvent swelling, and coal-solvent solubility have also been explored. While considerable progress has accompanied improvements in computational power and software advances, it is the generation of the model that is the most significant barrier to the meaningful utility of these models. The ability to generate large-scale models (incorporation of molecular weight diversity and structural diversity) with new automation approaches, coupled with new dynamic force-fields that can simulate reactive events in complicated materials like coals, offers a new hope for the utility of coal or char molecular models to probe our understanding and aid in the scientific method rather than our current informed trial and error approach. (C) 2010 Elsevier B.V. All rights reserved.