This article intends to provide an efficient modeling and compensation method for the synthetic geometric errors of large machine tools. Analytical and experimental examinations were carried out on a large gantry-type machine tool to study the spatial geometric error distribution within the machine workspace. The result shows that the position accuracy of the tool-tip is affected by all the translational axes synchronously, and the position error curve shape is non-linear and irregular. Moreover, the angular error combined with Abbe’s offset during the motion of a translational axis would cause Abbe’s error and generate significant influence on the spatial positioning accuracy. In order to identify the combined effect of the individual error component on the tool-tip position accuracy, a synthetic geometric error model is established for the gantry-type machine tool. Also, an automatic modeling algorithm is proposed to approximate the geometric error parameters based on moving least squares in combination with Chebyshev polynomials, and it could approximate the irregular geometric error curves with high-order continuity and consistency with a low-order basis function. Then, to implement real-time error compensation on large machine tools, an intelligent compensation system is developed based on the fast Ethernet data interaction technique and external machine origin shift, and experiment validations on the gantry-type machine tool showed that the position accuracy could be improved by 90% and the machining precision could be improved by 85% after error compensation.