Using the output from the run of GRAPES Meso for the six venues during the Vancouver 2010 Winter Olympic and Paralympic Games, its capability to forecast the 2 m temperature, relative humidity, 10 m wind speed and wind direction is evaluated in terms of the forecast accuracy, mean error, mean absolute error and Alpha Index. And also statistics TS, EH, PO, NH, B and ETS are used for the verification of precipitation. The results show that the forecast accuracy of relative humidity is the highest and the forecast performance tends to be steady with the increasing leading time. The model exhibits a cold bias for temperature forecast. As for relative humidity, the forecast experiences a gradual transition from dry bias to wet bias, while wind speed forecast is always larger than observation. Among all the classification for the precipitation, the forecasts of GRAPES Meso for rain or shine have the highest mean TS. And with the increase of precipitation, ETS is gradually close to the TS. At last compared with the other models we found that there are some deficiencies in GRAPES Meso. Through the verification it is found that there is a certain systematic error in the forecasts, and if the error can be effectively reduced then the numerical weather prediction will have a greater improvement.