Wheat is one of the major staple food crops around the world. Genetic dissection of important traits for disease resistance, adaptation and grain quality is essential to the continuous improvement of wheat yield and quality to meet the demand of global growing population. We conducted a large-scale genome-wide association study (GWAS) using a panel of 768 wheat cultivars that were genotyped by genotyping-by-sequencing and detected 548 quantitative trait loci (QTLs) for 12 important traits including grain yield and quality, resistance to stripe rust, leaf rust and powdery mildew, and cold tolerance. About 70% of the QTLs were delimited to ≤ 1.0 Mb intervals and eight of them are known genes that have been cloned. Three of the QTLs enhancing spike seed-setting and grain size were validated in three bi-parental populations and candidate genes were identified. Additive effects observed among the QTLs for most traits indicated that the phenotypes of these identified QTLs were highly predictable. Genomic prediction models for each trait were constructed based on the significant SNPs that were highly associated with those traits with about 80% prediction accuracies. Results from this study demonstrated that GWAS using a large population and high density marker coverage can not only lay a solid foundation for large-scale QTL fine mapping and candidate gene identification, but also increase the prediction accuracy for genomic selection in wheat.