This study estimated the genetic parameters of calving ease (CE) in Korean Holstein using linear animal-maternal (AMAT) and linear sire-maternal grandsire (SMGS) models. Calves born from the first three parities of cows (P1, P2, P3) between 2000 and 2024 were analyzed in two parity-level data subsets. The first subset comprised 133,998 (P1), 185,988 (P2), and 122,297 (P3) records. The second subset had at least seven records per herd-year subclass, with 104,469, 104,095, and 46,280 records for P1, P2, and P3, respectively. CE was defined as a calf trait, and the scores ranged between 1 and 4. Higher scores indicated greater difficulty at birth. Parity-level (co) variances were obtained for each dataset using the BLUPF90+ software package. Heritability (h2) values for direct effects ranged between 0.002 and 0.008. Maternal h2 values from the AMAT and SMGS models were between 0.002 and 0.353 and between 0.004 and 0.008, respectively. Genetic correlations between direct and maternal effects varied widely in the AMAT model but were relatively narrow in the SMGS model. The correlation of estimated breeding value (EBV) of sire between datasets and sire EBV reliabilities was more stable for SMGS than AMAT. We conclude that the AMAT model would be suitable for routine evaluations due to extensive population coverage, whereas SMGS would be better for robust genetic parameter estimations. To leverage the strengths of both models, we suggest using the genetic (co)variance components estimated from the SMGS model within the framework of the AMAT model for the national evaluation of CE in Korean Holstein cattle.