The objective of this study was to estimate genetic parameters of 39 novel mid-infrared predicted milk traits (e.g. nutritional quality, technological properties, metabolic status, environmental fingerprint) for three dual purpose cattle breeds (i.e. Dual-Purpose Belgian Blue (dpBB), Montbéliarde (MON) and Normande (NOR)), which are also used in organic farming in the Walloon Region of Belgium, as part of the 2-Org-Cows project. Edited data included 21,287, 10,062 and 4,637 first-lactation test-day records collected in the Walloon region of Belgium from 2,988, 1,330 and 621 dpBB, MON and NOR cows, respectively. Genetic parameters were estimated using REML applied to single-trait random regression test-day models for six conventional traits (yields, contents and somatic cell score) and the 39 novel mid-infrared predicted milk traits. Results for conventional traits allowed comparison to literature showing values that were close to the expected ones. For novel traits, comparison with available literature values for Holstein breed showed generally similar estimated heritabilities. Reported average daily heritabilities estimated for the 39 novel traits tended to be higher for dpBB (0.13-0.64) than MON and NOR (0.03-0.60) breeds. Few novel traits showed large differences between breeds except between dpBB and NOR for milk composition traits. However, results for NOR breed have to be taken very carefully given the low number of animals. Even if the used methane prediction equation was not yet validated for these breeds, estimated average daily heritability was moderately high for dpBB (0.41) and MON (0.36) and moderate for NOR (0.23) indicating that this prediction might also be useful in these dual purpose breeds.