Gross primary productivity (GPP) is an important parameter in describing terrestrial ecosystem productivity. This review surveys the existing remote sensing GPP estimation algorithms including vegetation index based and light use efficiency based models and their accuracies, and summarizes two 1 km spatial resolution GPP product accuracy under eight different vegetation types. MOD17, which is the most commonly used GPP product, provides global-scale spatio-temporal continuous data. A strong correlation exists between global-scale MODIS/GPP and in-situ measurement (R~2=0.59) with medium estimation accuracy (RMSE=2.86 gC/m~2/day). Estimation accuracy is high in deciduous broadleaved and evergreen coniferous forests but low in evergreen broadleaved forests and savanna. Finally, we analyze the uncertainties in GPP estimation and verification with the remote sensing method and suggest possible approaches to improve the accuracy of GPP estimation and its development tendency.