Various observations are made to invert for the source model of the 2015 Gorkha earthquake. Previously published slip models involve different slip distributions and slip amplitudes. To obtain a robust model, we jointly invert for the rupture process of the Gorkha earthquake using three types of observations: Global Positioning System (GPS), Interferometric Synthetic Aperture Radar (InSAR), and teleseismic P wave data. Three constraints (i.e., spatial and temporal smoothness and minimum moment constraints) are applied to regulate the rupture process. The weight ratios between data sets and prior constraints are determined from a new developed generalized Akaike’s Bayesian Information Criterion (gABIC) weight determination method. This gABIC method can simultaneously and objectively weigh various kinds of observations and prior constraints, rather than being limited to the two types in the traditional ABIC method. The inverted slip model reveals the seismic moment of the earthquake is 9.40×1020Nm released primarily in the first 60s, equivalent to Mw 7.92. The rupture front propagates unilaterally along the strike direction and spreads in both the strike and dip directions. The slip distribution shows one large slip patch with two peaks along the strike direction with a maximum slip amplitude of approximately 7.6m and identifies large slips (> 2m) that occur in the updip section of the Main Himalayan Thrust (MHT) fault plane. Our results indicate that the released slip amplitude is larger than the accumulated slip amplitude since 1833, so the 2015 earthquake likely does not overlap with the 1833 earthquake.