Information on the elemental abundances and distribution is essential for understanding the petrological characteristics and geological evolution of the Moon. In this paper, the thermal infrared data acquired by Lunar Reconnaissance Orbiter (LRO) Diviner are processed to investigate lunar elemental abundances on a global scale (60°N/S) for the first time. The Diviner Level 3 Standard Christiansen feature (CF) product with the resolution of 128 pixels/degree and the coverage of 99.86% is first analyzed and used. The Diviner global models are then established by the univariate regression methods based on the relationships between Diviner CF and ground truths of elemental abundances at 48 lunar sampling sites and the limitations of 1 RMSE of 48 datasets. Finally, the best maps of SiO, TiO, AlO, FeO, MgO and CaO abundances considering both resolution and coverage simultaneously are presented and analyzed from global, geologic units, crater and ejecta surfaces. The comparisons indicate that a satisfactory consistency is observed between Diviner results and Clementine or Chang’E (CE)-1 results, while Diviner results exhibit better practicability in presenting detailed information for elemental abundances on lunar surfaces and higher accuracy on the surface with high latitudes or poor light conditions. Meanwhile, it is also demonstrated that Diviner results is the reliable data sources for the applications in classifications of mare basalt, inhomogeneity of highland crust and Mg#.