Asphalt mastic, composed of asphalt binder and mineral filler, plays a critical role in determining the mechanical behavior and performance of asphalt mixtures. Conventional laboratory tests primarily evaluate the overall mechanical response of mastics and cannot capture their local behavior at the microstructural scale. This limitation highlights the need for numerical models that represent the heterogeneous microstructure of asphalt mastic. A key step in such modeling is generating representative microstructural geometries that reflect the characteristics of real samples. Since mastic behavior depends on both binder properties and filler characteristics, including gradation and volumetric content, this study developed specialized software to generate three-dimensional random particle geometries based on predefined gradation curves and volume fractions. Using this tool, two-phase numerical models were constructed according to the specifications of actual mastic samples. The viscoelastic properties of two asphalt binder types, a fresh binder and a reclaimed aged binder, along with their corresponding mastics, were determined through frequency sweep tests using a Dynamic Shear Rheometer. These experimentally measured properties were incorporated into the generated microstructures to develop finite element models of asphalt mastics. To validate the proposed approach, numerical simulations of frequency sweep tests were conducted at selected temperatures and loading frequencies, and the results were compared with experimental measurements. The comparisons demonstrated that the randomly generated geometries effectively represent mastic microstructures, and the developed models predicted viscoelastic behavior with high accuracy, achieving average agreements over 95% for complex shear modulus (G*) and over 90% for phase angle (δ).