Rice (Oryza sativa L.) is a globally important food source providing carbohydrates, amino acids, and dietary fiber for humans and livestock. The branched-chain amino acid (BCAA) level is a complex trait related to the nutrient quality of rice. However, the genetic mechanism underlying the BCAA (valine, leucine, and isoleucine) accumulation in rice grains remains largely unclear. In this study, the grain BCAA contents and 239,055 SNPs of a diverse panel containing 422 rice accessions were adopted to perform a genome-wide association study (GWAS) using a recently proposed 3VmrMLM model. A total of 357 BCAA-content-associated main-effect quantitative trait nucleotides (QTNs) were identified from 15 datasets (12 BCAA content datasets and 3 BLUP datasets of BCAA). Furthermore, the allelic variation of two novel candidate genes, LOC_Os01g52530 and LOC_Os06g15420, responsible for the isoleucine (Ile) content alteration were identified. To reveal the genetic basis of the potential interactions between the gene and environmental factor, 53 QTN-by-environment interactions (QEIs) were detected using the 3VmrMLM model. The LOC_Os03g24460, LOC_Os01g55590, and LOC_Os12g31820 were considered as the candidate genes potentially contributing to the valine (Val), leucine (Leu), and isoleucine (Ile) accumulations, respectively. Additionally, 10 QTN-by-QTN interactions (QQIs) were detected using the 3VmrMLM model, which were putative gene-by-gene interactions related to the Leu and Ile contents. Taken together, these findings suggest that the implementation of the 3VmrMLM model in a GWAS may provide new insights into the deeper understanding of BCAA accumulation in rice grains. The identified QTNs/QEIs/QQIs serve as potential targets for the genetic improvement of rice with high BCAA levels.