The principle of isostasy is important in different fields of geosciences. Using an isostatic hypothesis for estimating the crustal thickness suffers from the more or less incomplete isostatic model and that the observed gravity anomaly is not only generated by the topographic/isostatic signal but also by non-isostatic effects (NIEs). In most applications of isostatic models the NIEs are disregarded. In this paper, we study how some isostatic models related with Vening Meinez's isostatic hypothesis can be improved by considering the NIE. The isostatic gravity anomaly needs a correction for the NIEs, which varies from as much as 494 mGal to − 308 mGal. The result shows that by adding this correction the global crustal thickness estimate improves about 50% with respect to the global model CRUST2.0, i.e. the root mean square differences of the crustal thickness of the best Vening Meinesz type and CRUST2.0 models are 6.9 and 3.2 km before and after improvement, respectively. As a result, a new global model of crustal thickness using Vening Meinesz and CRUST2.0 models is generated. A comparison with an independent African crustal depth model shows an improvement of the new model by 6.8 km vs. CRUST2.0 (i.e. rms differences of 3.0 and 9.8 km, respectively). A comparison between oceanic lithosphere age and the NIEs is discussed in this study, too. One application of this study can be to improve crustal depth in areas where CRUST2.0 data are sparse and bad and to densify the resolution vs. the CRUST2.0 model. Other applications can be used to infer the viscosity of the mantle from the NIEs signal to study various locations around the Earth for understanding complete, over- and under-compensations of the topography.