An uncertainty index (from the uncertainty in measurement, proposed by the International Organization for Standardization) and the instability index (a derived index) are defined to evaluate the forecast performance of Numerical Weather Prediction (NWP) models. The uncertainty index is defined by comparing two mosaicked model fields with same meteorological variables from analysis and prediction, respectively. An operational system based on the two indexes is developed to evaluate the prediction performance for three operational NWP models (ECMWF, JMA and T213) in real time. Results from the operational application of these indexes show that these models can provide objective criteria for forecasters to evaluate NWP models, and be helpful for better use of NWP products.