The aim of the present work was to assess the model capabilities in describing the proton induced spallation reactions. To understand the reaction mechanism, di erent sets of observables must be investigated, inclusive as well as exclusive. Based on these criteria, the e orts were done to survey the scienti c literature for the selection of representative data sets which ts the need. The selected data were rich in terms of production of various ejectiles: neutrons, light charged particles (LCP: p, d, t, 3He, 4He), intermediate mass fragments, i.e., the particles with atomic mass number (ALCP < AIMF < Afissionfragments), and target-like heavy residua. Several atomic nuclei from Al up to Pb were selected as representative for all the targets. The proton beam covered the broad range of energies from 180 MeV to 3000 MeV. The spallation reaction was treated as a two stage process. In the rst stage, the incident proton initiates the cascade of binary collisions with target nucleons leaving behind an excited remnant. The second stage consists in the decay of this excited remnant nucleus. The selection of best models to describe each of these two stages was done on the basis of previous benchmark e orts where INCL4.5 model was found to be the best to describe the rst stage of the reaction. Therefore the newest version of this model - INCL4.6 was used in the present study. Four theoretical models di erent in approach to the reaction mechanism were chosen to realize the description of the second stage: ABLA07, GEMINI++, SMM, GEM2. Qualitative as well as quantitative comparisons of model calculations with experimental data were undertaken. To judge the quality of models the agreement in magnitude of di erent observables with model predictions as well as reproduction of the shape of the mass, angle and energy distributions of the cross sections were taken into account. Various deviation factors were used for providing ranking and validation of the spallation models. The statistical properties of the test factors, i.e., their expectation value, variance and probability density function were studied carefully. Two new statistical deviation factors named M􀀀factor and A􀀀factor were proposed in the present work. They are equally good as the best factors used up to now in the literature but are more intuitive. The ranking of models obtained by the application of the deviation factors were compared with the qualitative estimation of the data reproduction. It was found that all the studied models are able to reproduce the main characteristics of the data, however systematic deviations were observed and their interpretation was proposed.