Geology differentiation converts physical property models derived from geophysical data to a 3D quasi-geology model. It represents a step change in quantitative interpretation of geophysical data. However, quantifying the uncertainties of the differentiated geological units in a 3D quasi-geology model has been largely unexplored. We have developed an empirical method to construct 3D probabilistic quasi-geology models in the deterministic inversion framework. We used mixed Lp norm joint inversion to recover a large sequence of physical property models based on multiple airborne geophysical data sets. Prior petrophysical measurements were used to determine the acceptance and rejection of these models. We then performed geology differentiation for all these accepted models and obtained a set of 3D quasi-geology models, based on which we constructed probabilistic 3D quasi-geology models. Our work has broad implications for 3D geological model building based on multiple geophysical and/or petrophysical measurements.