SUMMARY.Dataset from Canadian and Uruguayan training populations were joined to analyse improvement of predictability for RFI. Three training populations where defined, URY (only data from Uruguay, 731), CAN (only data from Canada, 1168) and TOTAL (joint dataset, 1899). Genealogical information from the two countries was merged based on the international identification and cross reference list, with the pedigree file resulting in 17289 animals.The demands for livestock products are increasing, and beef production seems not to be an exception. This implies a challenge to beef production that has to increase productivity without increasing area or environmental footprint (a finite commodity), increasing costs (competing in disadvantage with chicken and pigs) or lowering product quality (its main advantage). The objective of present study was to compare the accuracy of genomic predictions for RFI based on national and bi-national training populations.