An evaluation of the predictive ability of a disturbed state constitutive model for softening behaviour of sand is performed. The model is shown to capture the softening behaviour of sand under both drained and undrained monotonic loading conditions. To use a particular constitutive model into a finite element or finite difference numerical procedure to solve real geotechnical problems, it is often desirable and necessary to include the variability in constitutive model parameters to capture the effects of model imperfections and inherent soil variability. Uncertainty in stress strain characteristics of sand due to inherent soil variability and model parameters’ variability are analysed. The sensitivity or the effects of disturbed state constitutive model parameters on stress dilatancy characteristics is stochastically analysed using first order second moment method to include the effects of model error and inherent soil variability. The most influencing parameters for both the deviatoric stress and volumetric strain are identified.