A multi-scale soil moisture and temperature monitoring network, consisting of 55 soil moisture and temperature measurement stations, has been established in central Tibetan Plateau (TP). In this study, the station-averaged surface soil moisture data from the network are used to evaluate four soil moisture products retrieved from the Advanced Microwave Scanning Radiometer-Earth Observing System (AMSR-E) and four land surface modeling products from the Global Land Data Assimilation System (GLDAS). Major findings are (1) none of the four AMSR-E products provides reliable estimates in the unfrozen season, in terms of the mission requirement of the root mean square error (RMSE) < 0.06 m3m−3. These algorithms either evidently overestimate soil moisture or obviously underestimate it, although some of them showed the soil moisture dynamic range, indicating that the retrieval algorithms have much space to be improved for the cold semi-arid regions. (2) The four GLDAS models tend to systematically underestimate the surface soil moisture (0–5 cm) while well simulate the soil moisture for 20–40 cm layer. In comparison with the satellite surface soil moisture products, three among the four models give low RMSE and BIAS values, but still falling out of the acceptable range. The causes for the modeling biases in this cold region were discussed.