Gross primary productivity (GPP) of terrestrial ecosystem is an important variable in studies of climate change and carbon cycle. Accurate estimation of GPP is crucial to understand ecosystem response to climate variability and change. The eddy covariance (EC) technique provides the best approach to measure net carbon dioxide (CO_2) change at site scale, which can be employed in GPP calculation. However, the EC technique only provides integrated CO_2 flux measurements over footprints with sizes and shapes that vary with the tower height, canopy physical characteristics and wind velocity. Satellite remote sensing has played an increasing role in the characterization of vegetation structure and the estimation of GPP, because it can overcome the lack of extensive flux tower observations in large areas. Among all the predictive methods, the light use efficiency (LUE) model may have the most potential to adequately address the spatial and temporal dynamics of GPP because of its theoretical basis and practicality, such as the vegetation photosynthesis model (VPM) and photosynthesis (PSN) model. The VPM is based on remote sensing and eddy covariance data, and has been validated in 21 sites of 10 kinds of terrestrial ecosystems. The PSN model is utilized to calculate the GPP and net primary productivity product, called MOD17, based on MODIS images provided by Goddard Space Flight Centre, National Aeronautics and Space Administration (NASA) of the USA. The VPM and PSN model have been widely used in the world, however little is known about the differences of the simulation results between these models. In this study, we used the maximum light use efficiency of flux site and PSN model to run the VPM (called FLUX-VPM and PSN-VPM), respectively. Then, we compared the simulation results of FLUX-VPM, PSN-VPM and MOD17 product with flux observation GPP (called Obs-GPP). Results showed that in Yucheng station where winter wheat/maize rotation was adopted and Yingke station where maize was planted, the correlation coefficient (r), root mean square error (RMSE) and modeling efficiency (EF) between GPP estimated by FLUX-VPM and Obs-GPP were 0.92, 1.30 g/(m~2·a), 0.84 and 0.97, 1.13 g/(m~2·a), 0.91, respectively, and GPP values were overestimated by 3.8% and 12.1%, respectively. The r, RMSE and EF between GPP estimated by PSN-VPM and Obs-GPP were 0.91, 2.66 g/(m~2·a), 0.31 and 0.96, 3.30 g/(m~2·a), 0.27, respectively, and GPP values were underestimated by 59.9% and 52.8%, respectively. The r, RMSE and EF between GPP from MOD17 product and Obs-GPP were 0.90, 2.87 g/(m~2·a), 0.32 and 0.97, 2.75 g/(m~2·a), 0.49, respectively, and GPP values were underestimated by 54.3% and 63.0%, respectively.