Light-use efficiency (LUE) is at the core of mechanistic modeling of global gross primary production (GPP). However, most LUE estimates in global models are satellite based and coarsely measured with emphasis on environmental variables. Others are from eddy covariance towers with much greater spatial and temporal data quality and emphasis on mechanistic processes, but in a limited number of sites. In this study, we conducted a comprehensive global study of tower-based LUE from 237 FLUXNET towers, and scaled up LUEs from in situ tower level to global biome level. We integrated the tower-based LUE estimates with key environmental and biological variables at 0.5°90.5° grid-cell resolutions, using a random forest regression (RFR) approach. Then, we developed a RFR-LUE-GPP model using the grid-cell LUE data. In order to calibrate the LUE model, we developed a data-driven RFR-GPP model using RFR method only. Our results showed LUE varies largely with latitude. We estimated a global area-weighted average of LUE at 1.23 ± 0.03 g C·m-2 ·MJ-1 APAR, which led to an estimate of global GPP of 107.5 ± 2.5 Gt C/yr from 2001 to 2005. Large uncertainties existed in GPP estimations over sparsely vegetated areas covered by savannas and woody savannas at middle to low latitude (i.e., 20°S–40°S and 5°N–40°N) due to the lack of available data. Model results were improved by incorporating Köppen climate types to represent climate/meteorological information in machine-learning modeling. This brought a new understanding to the recognized problem of climate dependence of spring onset of photosynthesis and the challenges in accurately modeling the biome GPP of evergreen broadleaf forests (EBF). The divergent responses of GPP to temperature and precipitation at middle to high latitudes and at middle to low latitudes echo the necessity of modeling GPP separately by latitudes.