A critical comparison of a hard-sphere discrete particle model, a two-fluid model with kinetic theory closure equations and experiments performed in a pseudo-two-dimensional gas-fluidised bed is made. Bubble patterns, time-averaged particle distributions and bed expansion dynamics measured with a nonintrusive digital image analysis technique are compared to simulation results obtained at three different fluidisation velocities. For both CFD models, the simulated flow fields and granular temperature profiles are compared. The effects of grid refinement, particle–wall interaction, long-term particle contacts, particle rotation and gas–particle drag are studied. The mechanical energy balance for the suspended particles is introduced, and the energy household for both CFD models is compared. The most critical comparison between experiments and model results is given by analysis of the bed expansion dynamics. Though both models predict the right fluidisation regime and trends in bubble sizes and bed expansion, the predicted bed expansion dynamics differ significantly from the experimental results. Alternative gas–particle drag models result in significantly different bed dynamics, but the gap between model and experimental results cannot be closed. In comparison with the experimental results, the discrete particle model gives superior resemblance. The main difference between both CFD models is caused by the neglect of particle rotation in the kinetic theory closure equations embedded in the two-fluid model. Energy balance analysis demonstrates that over 80% of the total energy is dissipated by sliding friction. Introduction of an effective restitution coefficient that incorporates the additional dissipation due to frictional interactions significantly improves the agreement between both models.