The aim of the present paper is to validate Lagrangian and Eulerian modeling approaches of biomass fast pyrolysis from comparison with experimental measurements. Wood samples are submitted during measured times to a controlled and concentrated radiation delivered by an image furnace. The heat flux densities are close to those encountered when wood is surrounded by hot bed particles in a dual fluidized bed (DFB) gasifier. In the image furnace, the sample is placed inside a transparent quartz reactor fed by a cold carrier gas. The volatile matter (condensable vapors and gases) released by the solid is quenched inside the reactor. It is hence possible to selectively study primary pyrolysis phenomena occurring at the solid level. All the pyrolysis products (char, vapors, and gases) are recovered, and their masses are measured as a function of the flash time allowing the assessment of mass balances. The yield of vapors does not significantly depend on the available heat flux density, unlike the gases and char yields. The experimental results are compared to data derived from two different modeling approaches. Their basic assumptions are discussed from characteristic time values which reveal the controlling phenomena. Mass transfer limitations are neglected in comparison with heat transfer and chemical phenomena. The first type of pyrolysis model relies on an original Lagrangian approach where mathematical equations of heat and mass balances are written with the assumptions that wood and char form two distinct layers. In the second one, a classical Eulerian approach is considered: equations are directly written at the whole particle level. The results of the two models as well as the experimental data (sample mass losses and product yields) are in quite good agreement.