Vessel flow quantification by two-dimensional (2D) phase-contrast magnetic resonance imaging (PC-MRI) using a three-dimensional (3D) magnetic resonance angiography (MRA) model to measure cerebral blood flow has unclear analytical reliability. The present study aimed to determine the inter- and intra-rater reliability of quantitative vessel-flow PC-MRI and potential factors influencing its consistency. We prospectively recruited 30 Asian participants (aged 20-90 years; 16 women; 22 healthy and 8 stroke patients) for performing 1.5-T MR equipped with a head coil. Each participant was first scanned for time-of-flight magnetic resonance angiography (TOF-MRA) images for localization of intracranial arteries. The 2D PC-MRI for each cerebral artery (total 13 arteries in fixed order) was performed twice by two well-trained operators in optimal position. Using the same 3D MRA as a map and facilitated with the non-invasive optimal vessel analysis (NOVA) system, each scan was taken on a plane perpendicular to the target artery. Two consecutive full 13-artery scans were performed at least 15 min apart after participants were removed from the scanner table and then repositioned. A total of four PC flow images obtained from each target artery were transmitted to a workstation facilitated with the NOVA system. Flow data were calculated semi-automatically by the NOVA system after a few simple steps. Two-way mixed-effect models and standard errors of measurements were used. In 13 cerebral arteries, repeatability, using the intra-rater estimate expressed as the average-measures intraclass correlation coefficient, ranged from 0.641 to 0.954, and reproducibility, using the inter-rater estimate, ranged from 0.672 to 0.977. Except in the middle cerebral artery and the distal segment of the anterior cerebral artery, repeatability and reproducibility were excellent (intraclass correlation coefficient exceeded 0.8). The use of quantitative vessel-flow PC-MRI is a precise means to measure blood flow in most target cerebral arteries. This was evidenced by inter-rater and intra-rater correlations that were good/excellent, indicating good reproducibility and repeatability.