Abstract: In order to make better use of a recombination inbred line (RIL) population derived from the cross of Shuhui 527 and Mianhui 436 in breeding, the genome-wide selection (GS) model for this population was constructed, and the optimal model was evaluated for the future screening of superior restorer lines to improve the breeding efficiency. Three GS models were constructed after detecting the agronomic traits and genotypes of the population quantity. Then, by evaluating the performance of GS model on various traits of the RIL population, the accuracy and predictive ability of the three prediction models applied in this population were explored. The results showed that: the eight traits such as the ear sprouting period, effective tiller number, plant height, grain length, grain width, panicle length, grain number per panicle and yield per plant of rice RIL population were all the quantitative traits controlled by polygenes genetically, making them suitable for GS analysis. The genotypes of the population were obtained by high-throughput sequencing. After screening and filtering, 454 129 SNPs from 181 accessions were obtained and three predictive models were constructed. The comprehensive evaluation based on the model accuracy and predictive capacity revealed that the GBLUP model exhibited superior overall performance, demonstrating its potential for practical implementation in the subsequent use.