Gravity field modeling using airborne vertical component gravimetry has made significant strides over the last decade. We demonstrate the feasibility of extending this to three-dimensions using data from inertial navigation systems (INS) and the Global Positioning System (GPS). A significant advantage of measuring the horizontal gravity components is that the geoid can be determined in profiles by direct along-track integration, thus not only adding strength to conventional methods, but reducing the required area of survey support, especially along model boundaries. As such, the ultimate limitation of the method is in the quality of the INS and GPS sensors. In our test case, all three components of the gravity vector were determined over a profile in the Canadian Rocky Mountains. Differences between available truth data and the computed gravity components have standard deviations of 7--8 mGal (horizontal) and 3 mGal (vertical). These standard deviations include uncertainties in the truth data (<5 mGal, for horizontal; 1.3 mGal, for vertical). The resolution in the computed values is about 10 km. These analyses have demonstrated for the first time that the total gravity vector can be determined from airborne INS and GPS to reasonable accuracy and resolution, without any external orientation information, nor prior statistical hypothesis on the gravity signature, using medium-accuracy INS and geodetic quality GPS receivers. ¿ 1999 American Geophysical Union