Accurate ocean tide models are required to remove tidal loading effects in geophysical research. Beyond a mere intercomparison, the accuracy of eight modern global models (DTU10, EOT20, FES2014b, FES2022b, GOT4.10c, HAMTIDE11a, OSU12, TPXO10-atlas-v2) and one regional model (NAO99Jb) was assessed in the eastern China marginal seas (ECMSs) using geodetic measurements. This involved rigorous comparisons with the tidal constant measurements at 65 tide gauges and with the GPS-measured M2 vertical ocean tide loading (OTL) displacements at 22 sites. The selected models showed significant disagreements close to the coasts of eastern China and the western Korean Peninsula, where the largest discrepancy for the M2 constituent could exceed 30 cm. However, EOT20 and FES2014b provided relatively close results, differing by only about 15 cm in Hangzhou Bay. EOT20 compared more favourably than the others to the tidal constant measurements, with a root sum square (RSS) of 11.1 cm, and to the GPS-measured M2 vertical OTL displacements, with a root mean square (RMS) of 0.49 mm. In addition, to differentiate between ocean tide models with subtle discrepancies when comparing them with the OTL measurements, consideration of the asthenospheric anelasticity effect was necessary.