Genomic selection is a promising new tool for plant breeders. So far, genomic selection has mainly been studied using single-trait models. However, a range of traits is typically phenotyped in breeding programs. Genetically correlated traits might be combined in multi-trait models to improve predictive abilities. Advanced wheat breeding lines from the Danish plant breeding company Nordic Seed A/S were phenotyped for the quality traits grain protein content, thousand-kernel weight, falling number and Zeleny sedimentation. The lines were genotyped using a 15k SNP array, and single-trait and multi-trait genomic selection models were evaluated using different cross-validation strategies. Predictive abilities ranged from 0.13 – 0.49, when predicting across breeding cycles. The predictive ability was lowest for grain protein content. However, predictive ability increased significantly when phenotypic data for the other traits were included for lines in the validation set using a multi-trait model. Thus, implementing genomic selection, including multi-trait models, in wheat breeding programs will improve selection of wheat lines with enhanced quality traits.