A virtual calibration chamber was developed using a three-dimensional (3D) discrete element method (DEM) to perform cone penetration tests (CPTs) on a discrete analogue of Ticino sand. The macroscale response of the DEM model was previously shown to be in good quantitative agreement with that of analogous physical models. In the current study the performance of the model at meso and microscale levels of resolution is examined. The microscale response is examined using particle displacements and contact force distributions. The mesoscale behaviour is examined using stress and strain fields obtained through appropriate averaging and interpolating procedures. Four CPTs are examined at the steady-state penetration stage. The effects of radial boundary conditions, initial stress state, initial average density, and particle rotational inertia are examined. The ability of the micro and mesoscale data to identify and explain the relevant mechanisms underlying the significant differences in the macroscale response of the models is discussed. Comparisons with similar phenomena observed in physical tests are also highlighted.