The pyrolysis of biomass in a fluidized-bed reactor is studied by a combination of a CFD-DEM algorithm and a multistep kinetic scheme, where fluid dynamics, heat and mass transfer, particle collisions, and the detailed thermochemical conversion of biomass are all resolved. The integrated method is validated by experimental results available in literature and a considerable improvement in predicting the pyrolysis product yields is obtained as compared to previous works using a two-fluid model, especially the relative error in the predicted tar yield is reduced by more than 50%. Furthermore, the evolution of light gas, char and tar, as well as the particle conversion, which cannot easily be measured in experiments, are also revealed. Based on the proposed model, the influences of pyrolysis temperature and biomass particle size on the pyrolysis behavior in a fluidized-bed reactor are comprehensively studied. Numerical results show that the new algorithm clearly captures the dependence of char yield on pyrolysis temperature and the influence of heating rate on light gas and tar yields, which is not possible in simulations based on a simplified global pyrolysis model. It is found that, as the temperature rises from 500 to 700 °C, the light gas yield increases from 17% to 25% and char yield decreases from 22% to 14%. In addition, within the tested range of particle sizes (<1 mm), the impact on pyrolysis products from particle size is relatively small compared with that of the operating temperature. The simulations demonstrate the ability of a combined Lagrangian description of biomass particles and a multistep kinetic scheme to improve the prediction accuracy of fluidized-bed pyrolysis.