Verification of typhoon track forecast in 2007 was conducted to understand the p erformance of five operational typhoon numerical models of China. The datasets i nclude the typhoon center location data officially issued by CMA and t he associated prediction of typhoon track provided by five typhoon operational f orecast centers of China. The evaluation includes distance error, skill score an d index of stability, etc. The result shows that: 1) the minimum (maximum) case- averaged 24h/48h distance error, the minimum(maximum) 24h/48h distance error for the best(worst) case prediction and 24h/48h model-averaged distance error are 1 22.8km/246.3km (180.7km/304.4km), 11.2km/10.1km (1429.7km/1003.7km) and 147km/2 67km, respectively. Pabuk (No. 0707 numbered by CMA) brought big difficulty to a majority of typhoon models; 2) the model-averaged skill score against the clima tic-persistent statistical track prediction approach of Shanghai Typhoon Institu te (STI) is 32%/43% for 24h/48h prediction; 3) stability indexes in terms of dis tance, direction of movement and availability are also presented to give an over all evaluation. Significant advantage of numerical model is identified against t he traditional statistical method in typhoon track prediction. However, it is al so recognized that most of the typhoon models witness deficiency in the case of landfall, particularly when typhoon making abrupt turning. This suggests that, i n addition to vortex initialization, more attention should be paid on improving the physics parameterization of boundary layer and land surface in typhoon numer ical model.