Evaluation of scenarios has become a critical component of climate change research. The performance of various GCMs in simulating observed climate at several stations in the Upper Indus River Basin was evaluated using different statistical indicators. The HadCM3, SRNIES, ECHAM4, and NCAR models simulated the spatial variability of mean temperature well, but the precipitation simulation was found to be inadequate. The results indicate that all the models substantially underestimated the magnitude of temperature on a monthly basis. The projected precipitation, however, showed much higher inter-model variability compared to mean temperature at all stations. Overall, a considerable inter-model variability in the simulation of observed climate was found at all stations. Owing to its ability to simulate the observed climate well, the HadCM3 model was selected for projection of future climate in the Upper Indus River Basin under several emission scenarios. Using the HadCM3 model, the temperature and precipitation changes across the UIRB were determined for two time periods; 2021-2050 (F1), and 2061-2090 (F2), relative to 1961-1990. The projected mean annual temperature across the study area varied from 1.7 to 4.3°C (for F1 and F2). The HadCM3 model predicted higher increase in mean winter temperature (from 2.1 to 4.5°C) compared to other seasons, whereas the increase in precipitation was observed in all seasons except winter. The increase in annual precipitation of 16 to 28% was projected for the two future periods considered in the present study.