The study introduces two Bayesian methods, BayesCr and BayesDr, to address the limitations of BayesA and BayesB in genomic prediction. These methods treat the prior probability π that a SNP has zero effect as unknown, aiming to improve the accuracy of genomic estimated breeding values (GEBVs) and provide insights into the genetic architecture of traits. The methods were evaluated using simulated scenarios and real data from North American Holstein bulls. Results show that BayesCr estimates π more accurately than BayesDr, with π being sensitive to the number of simulated QTL and training data size. Milk yield and fat yield traits have larger-effect QTL compared to protein yield and somatic cell score. Despite the drawbacks of BayesA and BayesB, they did not impair the accuracy of GEBVs. The accuracies of alternative Bayesian methods were similar, and BayesA performed well for the given SNP density. Computing time was shorter for BayesCr compared to BayesDr, and the longest for the implementation of BayesA. The study concludes that BayesCr is a promising method for routine applications, considering its computational efficiency and ability to infer the number of QTL.The study introduces two Bayesian methods, BayesCr and BayesDr, to address the limitations of BayesA and BayesB in genomic prediction. These methods treat the prior probability π that a SNP has zero effect as unknown, aiming to improve the accuracy of genomic estimated breeding values (GEBVs) and provide insights into the genetic architecture of traits. The methods were evaluated using simulated scenarios and real data from North American Holstein bulls. Results show that BayesCr estimates π more accurately than BayesDr, with π being sensitive to the number of simulated QTL and training data size. Milk yield and fat yield traits have larger-effect QTL compared to protein yield and somatic cell score. Despite the drawbacks of BayesA and BayesB, they did not impair the accuracy of GEBVs. The accuracies of alternative Bayesian methods were similar, and BayesA performed well for the given SNP density. Computing time was shorter for BayesCr compared to BayesDr, and the longest for the implementation of BayesA. The study concludes that BayesCr is a promising method for routine applications, considering its computational efficiency and ability to infer the number of QTL.