Adoption of accurate and simplistic biometric models that estimate variance components, predicting genotypic effects are desired in plant breeding. The aim of this study was to estimate the variance components and predict genotypic values by REML/BLUP for combinations of maize inbred lines derived from two heterotic groups with F1's evaluation in four locations. The crosses originated 25 hybrids that were evaluated in four trials. Genotype x environment interaction provided oscillations regarding the best combinations in the environments and must be individually analyzed in each environment, the superior crosses for grain yield, except 15x3’ combination, which achieved good performance in all the trials. The combination of 5x4' is promising for obtaining high grain yield in Itapiranga and Pato Branco. The 5x3’ combination have a higher overall average, being higher than the other combinations in Clevelândia and Ampere environments for plot yield, providing hybrids with fewer branches in the tassel. Key words: Best linear unbiased predictor, restricted maximum likelihood, Zea mays L.