The Montreal Cognitive Assessment (MoCA) has been proved brief and sensitive to screen for mild cognitive impairment (MCI) and early dementia in some developed countries or areas. However, little MoCA data are available from mainland China. In this study, the MoCA was applied to 8411 Chinese community dwellers aged 65 or older (6283 = cognitively normal [CN], 1687 = MCI, and 441 = dementia). The MoCA norms were established considering significant influential factors. The optimal cutoff points were 13/14 for illiterate individuals, 19/20 for individuals with 1 to 6 years of education, and 24/25 for individuals with 7 or more years of education. With the optimal cutoffs, the sensitivity of the MoCA was 83.8% for all cognitive impairments, 80.5% for MCI and 96.9% for dementia, and the specificity for identifying CN was 82.5%. These indicate that with optimal cutoffs, the MoCA is valid to screen for cognitive impairment in elderly Chinese living in communities.