The aims of our study were to estimate genetic parameters for body weight and visual scores and to evaluate their inclusion as selection criteria in the Nelore breeding program in Brazil. The traits studied were the body weight adjusted to 210 (W210) and to 450 (W450) days of age and visual scores for body structure, finishing precocity, and muscling evaluated at weaning (BS, FP, and MS) and yearling (BS, FP, and MS) ages. A total of 33,242, 26,259, 23,075, and 26,057 observations were considered to analyze W210, W450, and visual scores at weaning and yearling. The significant (P < 0.05) fixed effects for all traits were farm, birth season, birth year, sex, and management group. Single-trait analyses were performed to define the most fitting model to our data using the average information restricted maximum likelihood algorithm, for weaning traits. Subsequently, these models were used in single- and two-trait analyses considering the Bayesian inference algorithm. Two-trait Bayesian analyses resulted in average direct heritability estimates for BS, FP, MS, W210, BS, FP, MS, and W450 of 0.28, 0.30, 0.27, 0.28, 0.40, 0.44, 0.39, and 0.50, respectively. Genetic correlations varied from 0.40 to 0.96. Benefits to animal performance can best be achieved by considering body structure, finishing precocity, and muscling as selection criteria in the Nelore breeding programs. The decision to use visual scores measured at weaning should be considered in order to decrease generation interval and assist pre-selecting individuals, expecting carcass improvements in the future progeny.