Abstract: Global/Regional Assimilation PrEdiction System (GRAPES) is an operational Numerical Weather Prediction (NWP) model which has been developed by Chinese scientists since 2000. This paper summarizes the history of development and the key technologies. To represent the uncertainty of initial conditions of GRAPES model, the Singular Vectors (SVs) has been applied to initial condition perturbation method of GRAPES Global Ensemble Prediction System (GRAPES-GEPS) and the ensemble transform of Kalman filter (ETKF) and ETKF-based multi-scale blending have been applied to initial condition perturbation methods of GRAPES Regional Ensemble Prediction System (GRAPES-REPS). To represent the uncertainty of GRAPES model itself, the stochastically perturbed parameterization tendencies (SPPT) and Stochastic Kinetic Energy Backscatter (SKEB) have been used to the both of GRAPES-GEPS and REPS. The challenges and future plans of GRAPES-GEPS and -REPS are analyzed.