The numerical modeling of solid pyrolysis for fire simulations consists of three stages: specifying the reaction scheme and physical models, estimating the kinetic and thermal parameters from experimental data, and then solving the system using a computer. The interpretation of the experimental input parameters must be verified by reproducing the experimental conditions with the same model for which the parameters are being sought. In this work, the performance of three previously proposed reaction schemes of wood pyrolysis in reproducing thermogravimetric experiments of birch wood was evaluated, and the remaining model parameters from small and bench scale calorimetric experiments were determined. The predictive capability was tested by cone calorimeter experiments at different heat fluxes. The results indicate that the first-order single-step reaction scheme can provide equally good predictions for the heat release rate as the more complex schemes. The source of the thermal parameters-direct measurement or inverse modeling-did not have a great influence on the predictive capability.