Based on the hindcast data of Beijing Climate Center (BCC) second-generation seasonal prediction model system (BCCv2), the seasonal prediction performance of 2-m temperature, precipitation and circulation was evaluated by employing deterministic and probabilistic forecast verification methods. BCCv2 simulations were compared with that of BCC first-generation prediction system (BCCv1) to further analyze the seasonal climate predictability. The results show that the performance of the BCCv2 is significantly improved compared to that of the BCCv1 especially in the tropical eastern Pacific Ocean, the Indian Ocean and the Maritime Continent areas. The improvements of precipitation prediction in the tropics are the major reason for the improvements of the forecast skill in the mid-high latitudes through realistic description of the atmospheric teleconnection patterns, such as the Pacific-North American (PNA) and East Asian -Pacific (EAP) patterns. The El Niño and South Oscillation (ENSO) signal is the dominant source of predictability for both the tropical and extra-tropical regions. The global atmosphere circulation in response to ENSO signal is accurately described in BCCv2, which improves its overall prediction performance by advancing ENSO prediction skill. From the perspective of probabilistic prediction, the BCCv2 showed useful prediction skills for the prediction of China surface air temperature anomalies in winter and precipitation anomalies in summer especially for the above normal (AN) and below normal (BN) events of winter temperature in eastern China with relatively high reliability and resolution. Therefore, further improvements of the capability of the BCCv2 in predicting tropical large-scale anomalies and primary climate variability modes and the application of probabilistic prediction products of this model are two key issues for improving seasonal climate prediction in China.