Crossbreeding is a growing strategy among dairy farmers, usually chosen to compensate weaknesses of a breed, in order to obtain robust cows, to facilitate the transition to a different farming system (organic farming,specifications of a PDO, etc.) or to combine various breed strengths. One of the objectives of the GenTORE project is the implementation of genomic evaluation on crossbred cows, which requires the imputation of crossbred genotypes with sufficient precision. In this study, we used genotypes from Normande × Holstein crossbreds, bred on the French experimental farm Le Pin au Haras. 1,208 crossbred animals with known pedigree were genotyped with the Illumina 50k SNP chip. Out of them, 243 were used as a validation population and their genotypes were reduced to the markers of the Illumina 10k SNP chip. The genotypes of the validation animals were imputed to 50k, using the other crossbred animals and the purebred ancestors as a training population. FImpute 2.2, Beagle 5, and Eagle 2.4.1 were used for imputation. Imputed genotypes and true genotypes of the validation population were compared and an imputation error rate was measured for each software. The three software were compared based on imputation error rate and memory requirements. FImpute always performed as good or better than the other two for all criteria and hence was determined the most suitable one for the imputation of crossbred animals. For FImpute, error rates were lower than 0.1%, probably partly thanks to the very small and very well known pedigree, and had low memory requirements. Hence, imputation of crossbred dairy cattle seems feasible for the purpose of routine genomic evaluations, even based on small datasets.