Genetic evaluation models is becoming more complex and take into account molecular information (SNP, QTL or major gene) to enhance selection of animals. Improvement of accuracy of genomic predictions on traits of interest in French dairy goats was a major issue. In this study, ssGBLUP, WssGBLUP methods and its alternative were compared and performed on five milk productions, five udder type and one functional traits. Data came from official genetic evaluation of January, 2012. Using WssGBLUP with a common weight for a window of 40 SNP that sums or averages the SNP variance enhances accuracy of genomic breeding values, especially for traits influenced by major gene or QTL. Differences on accuracy between Alpine and Saanen breeds, where an important QTL on chromosome 19 was only detected in Saanen for many traits, were highlighted with WssGBLUP method.