To investigate the role of the Montreal Cognitive Assessment (MoCA) (Beijing version) and its memory tasks on detecting different mild cognitive impairment (MCI) subtypes including amnesticMCI (aMCI) and nonamnestic MCI (naMCI) in memory clinics. A total of 121 patients with MCI and 53 healthy controls were included. Fifty-six aMCI-multiple domains (amMCI), 32 aMCI-single domain (asMCI), and 33 naMCI patients were diagnosed according to extensive cognitive tests. All participants were administered by the Mini Mental State Examination (MMSE) and the MoCA. Patients with amMCI performed worse than patients with asMCI, naMCI, and healthy controls on the MMSE and the MoCA (p < 0.001). The area under the curve (AUC) value for the MoCA when comparing the amMCI and control groups was 0.884 (p < 0.001), which was superior to that of the MMSE. The AUC value decreased to 0.687 when applied to the naMCI and control groups (p = 0.007), which was still higher than that of the Rey Auditory Verbal Learning Test (RAVLT) or the Rey-Osterrieth complex figure (ROCF). Delayed free recall or category prompted recall in the MoCA had roles in differentiating asMCI and controls groups with AUC value of 0.717 (p = 0.002) and 0.691 (p = 0.005), respectively. The MoCA is a good screening tool for detecting different types ofMCI and is suitable for patients in outpatient clinics.