An airborne gravity survey for Nepal was carried out in December 2010 with the primary goal to provide data for a new national geoid model, which will in turn provide gravity information for the future global gravity field EGM2020. This gravity data is used again to determine the regional geoid model (NPG20) for Nepal using Least-Squares Collocation (LSC) with Remove-Compute-Restore approach. In comparison to the previously computed geoid model, Nepal Geoid 2011 (NPG11) using EGM2008, this study applies XGM2019e as the global model. The comparative study shows, XGM2019e fits the airborne gravity observations much better than EGM2008 in the study area. The computation of geoid heights is done using LSC with the determination of proper covariance function for the gravity data, while the previous study includes a combination of LSC for downward continuation and spherical FFT for the calculation of the geoid. This contribution evaluates the benefit of our two main adaptions. The comparative study of geoids, NPG11 and NPG20 showed that there exist significant differences between these models especially in the area where the elevation is higher than 7000 m. The data analysis of the study showed that the currently available airborne gravity data was not sufficient to provide high frequency gravity signals and the significant differences in these geoid models was solely related to the different handling of the high-frequency gravity field component of the background model, i.e. EGM2008 and XGM2019e.