<p>This paper presents a contribution to integrating the stochastic behavior of sandy soils using the NORSAND constitutive model within the Random Finite Element Method (RFEM), extending its application beyond conventional deterministic analyses. Representative data for Erksak sand are first compiled from the literature, followed by a material-point sensitivity analysis to identify the key parameters governing drained and undrained behavior. Stochastic analyses are then performed using Monte Carlo simulations to quantify the probabilistic response in terms of pore-pressure buildup, deviatoric stress, and liquefaction potential. A further contribution of this study lies in the stochastic evaluation of the bearing capacity of a loose sandy slope benchmark problem using the NORSAND-based RFEM approach, incorporating anisotropic random fields and an uncoupled effective-stress-based hydro-mechanical formulation. The stochastic NORSAND-based framework enables the propagation of the state-dependent contractive response and post-peak softening under undrained conditions, leading to reduced bearing capacity levels, in contrast to simplified constitutive approaches. Furthermore, the impact of Bayesian-informed parameter statistics was assessed in this example and compared with a scenario characterized by higher parameter variability spanning the full contractive range of the sand. In this latter case, the computed stochastic bearing capacity follows a Gaussian distribution, ranging from approximately 0.85–1.2 times the deterministic prediction, whereas the use of Bayesian-informed parameters leads to a markedly reduced variability. These results quantify the role of geotechnical uncertainty on slope performance.</p>