The information on the projected climate changes over China is of great importance for preparing the nation’s societal adaptiveness to the future natural ecosystem. This study reports the surface mean temperature changes during 2014–2100 over China and its four sub-regions (Northern China, Northwestern China, Southern China, and the Tibetan Plateau) by analyzing 20 models from the Coupled Model Intercomparison Project Phase 6 (CMIP6) under three Shared Socio-economic Pathway (SSP) scenarios: SSP126, SSP245 and SSP585. The multi-model ensemble mean (MMEM) of 20 CMIP6 models has cold biases over China during 1979–2014, with improved performance compared with the CMIP5 models. In contrast, the CMIP6 models simulate well in the spatial climatology with lower warming rates over China. Relative to 1986–2005, the regionally averaged surface mean temperatures from the MMEM over China under SSP126, SSP245, SSP585 scenarios are projected to increase by 1.31 °C, 1.32 °C, 1.45 °C in the near-term (2021–2040), 1.75 °C, 2.06 °C, 2.66 °C in the mid-term (2041–2060), and 1.08 °C, 2.97 °C, 5.62 °C in the long-term (2081–2100), respectively. The CMIP6 models simulate accelerated warming occurs over the Northwestern China and the Tibetan Plateau, suggesting that the arid and semi-arid regions are particularly sensitive to future climate warming. We quantify uncertainty for future projections of temperature changes over China, and the main sources of uncertainty are model and scenario uncertainty particularly for the regions with the largest cold bias. This suggests that the observational constraints on these regions will lead to significant improvements for climatic projections over China.