Untangling the genetic architecture of grain yield and yield stability is a crucial determining factor to optimize genomics-assisted breeding strategies in wheat. We conducted haplotype-based genome wide association study on a large set of seven cohorts of advanced bread wheat lines (n= 6,518) within CIMMYT elite yield trials grown over seven years under contrasting environments (irrigated and stress) to identify important genomic regions linked to grain yield (GY) and the yield stability (superiority index Pi). Potential genomic regions were identified on chromosomes 1A, 1B, 2B, 3B, 4A, 4B, 5B, 6B and 7B that are consistent for stress environments or across multiple environments. Thus, we have identified an interesting repertoire of beneficial alleles for GY in CIMMYT germplasm with allelic effects ranging from 2 to 10%. Epistatic interactions contributed to an additional 5 to 9% variation on average. Lines have been identified with different combinations of beneficial alleles to be included in crossing and line development schemes. Further, we have explored whether integrating a subset of these consistent associations as fixed variables in prediction models improves prediction accuracy for GY and yield stability. For GY, the model led to up to 10% increase in prediction accuracy, whereas for Pi this approach did not provide any advantage. This is a step forward in understanding the genetic architecture of grain yield and yield stability in CIMMYT spring wheat germplasm and for its potential use in breeding.