Implementation of genomic prediction can bolster rates of genetic gain in sorghum improvement and permit more efficient allocation of resources within hybrid breeding programs. In the present study, alternative genomic prediction models were compared to assess the potential benefits of including inbred phenotypic records, dominance effects, and genotype-by-environment (G×E) interactions in predicting hybrid grain sorghum performance. Comparisons were made in a set of 395 hybrid combinations derived from 92 parental inbred lines tested in a sparse multienvironment trial. Phenotypic data were collected on hybrids and inbreds for days to mid-anthesis, grain yield, and plant height, and genomic data on parental inbreds were collected by genotyping × sequencing. A significant increase in prediction accuracy was observed when modeling G×E effects; however, dominance effects did not contribute to the overall predictive ability of models in this data set. Including phenotypic data from parental lines significantly improved the prediction of hybrid merit by as much as 17% for days to mid-anthesis, 14% for grain yield, and 33% for plant height when there were no testcross records for a given parental line. Alternatively, similar improvements were not as consistent when the training set included lines already tested in hybrid combinations. Thus, hybrid crop breeders can further optimize genomic predictions for un-testcrossed lines by including non-additive effects and inbred data. INTRODUCTIONGrain sorghum [Sorghum bicolor (L.)] is an ancient cereal grain that is the second most widely grown feed grain in the United States on a production basis (USDA, 2021) and ranks as one of the top 25 major commodities in the world since 1961 (Monk et al., 2015). As a drought-tolerant crop Abbreviations: BLUE, best linear unbiased estimator; BLUP, best linear unbiased predictor; CV, coefficient of variation; GBLUP, genomic best linear unbiased prediction; G×E, genotype-by-environment.