Digital gridded weather forecast is the developing trend of weather forecasting operation in China. Based on NWP model products, meteorological observation data and an evaluation on subjective and objective forecast, a diurnal extreme temperature Multimodel Consensus Gridded Forecast (McGF) system was developed. Statistical results show that there are significant seasonal forecast error differences in both subjective and objective forecasts, and forecasters have more forecasting skills in summer, when the NWP systemaitc errors are more stable. NWP model shows a flow dependent (conditional bias) characteristic. When the temperature is higher (lower), the nagative (positive) forecast error is bigger. Both subjective and objective forecasts are affected by topography and there are relatively significant forecast errors in the northern mountainous areas. As the lead time of forecast extends, the growth of forecast errors is, not big and the subjective forecasting skills are stable relatively. Based on these results, McGF interpretation application system was devloped with four modules, including realtime verification, stationbased interpretation, gridded application and performanceweighted averages. The results showed that Tmax forecasts of McGF are better than subjective and objective forecasts, with its mean absolute errors less than 2℃ within 72 h. Relatively speaking, Tmin forecast errors are much lower and the enhancements of McGF are relatively small. The cases of extreme high/low temperature showed that McGF gridded forecasts in Guangdong Province can more reasonably reflect the spatial distribution and intensity feature.