This paper introduces the use of a probabilistic inverse model method to calibrate numerical models based on Particle Flow Code in three dimensions (PFC 3D). The main aim is to fully define the parameters of the particle models through a joint probability density function conditioned on experimental observations obtained from triaxial rock testing. The discrete particle model built for the rock specimen simulation is comprised of an assembly of spheres of specified stiffness bonded together with specified strengths. A case study is presented where the probabilistic calibration method is applied to experimental results of triaxial compressive tests of Vosges sandstone specimens tested at the L3S-R (Grenoble, France). Calibration results will serve for simulating likely rock specimens' behaviour when matching not only the global stress-strain and volumetric responses of jointed rock for different confining pressures, but also the fracture modes observed in the laboratory. Additionally, probability descriptions of the models' performance at different levels of axial strain are also retrieved, enhancing the quantity and quality of inferences that are obtained traditionally by optimization methods for inverse problems. Preliminary results include in-progress trends and model relationships as the probabilistic calibration is currently being computed. © 2009 IOS Press.