Genetic parameters and selection responses were obtained for fillet weight and fillet yield of Progift Nile tilapia (Oreochromis niloticus) in China after six generations of multi-trait selection for growth and fillet yield. A total of 9619 test fish representing 687 full-sib families in six generations (G1–G6) of Nile tilapia originating from the GIFT breed were sacrificed to record skin-on fillet weights. Some of these skin-on fillets were further processed by skinning (5971 test fish) and trimming (4633 test fish) to allow calculations of three estimates of fillet yield (based on skin-on, skinned and trimmed fillets). Recorded fillet weights were positively influenced by body weight and negatively influenced by age at recording. The heritability (h2) of skin-on fillet weight showed large variation in magnitude between generations (range of 0.00–0.45), but was of medium magnitude (0.30) when analyzing all data combined. The h2 of fillet yield, which varied between 0.08 and 0.30 in different generations, was relatively stable (0.17–0.23) for different estimates of fillet yield when analyzing across all generations. Including all data, the effects common to full-sibs (c2) accounted for 8% and 1–2% of the total phenotypic variance, respectively, for skin-on fillet weight and different estimates of fillet yield. The genetic correlations between different estimates of fillet yield were all very high (0.95–0.97 when analyzing all data) showing that it is sufficient to select based on skin-on fillet yield. The genetic correlation between skin-on fillet weight and body weight at harvest was also very high (0.97), while that between fillet yield and fillet weight was of moderate magnitude (0.33). The genetic correlation between fillet yield and body weight at harvest was not significantly different from zero. Genetic trend analysis based on all data predicted accumulated selection responses of 121g (1.87 phenotypic standard deviation units) larger skin-on fillet weight and 1.2%-units higher skin-on fillet yield after six generations of multi-trait selection. It is concluded that the ongoing program in China has resulted in considerable genetic improvement of fillet weight, and that genetic changes of fillet yield is a much slower process (0.2%-units per generation). Implications for commercial selective breeding programs are discussed.