BACKGROUND: VO2 may be predicted with base on anthropometric and physiological parameters for determined populations. OBJECTIVE: To propose models for submaximal and maximal VO2 prediction in young Brazilian adults. METHODS: A total of 137 volunteers (92 men) underwent graded maximal exercise test (GXT) in a cycle ergometer (MonarkTM, Br). Gas exchange and respiratory measurements were performed in an open circuit (AerosportTM TEEM 100, USA). In another group, 13 volunteers underwent GXT and a square wave test (SWT) in order to evaluate the external validity of Neder et al’s formula, ACSM’s formula, and of Åstrand-Ryhming nomogram. The study design chosen was a cross-validation and the significance level was set at p < 0.05. RESULTS: For men during submaximal exercises, a mathematical model was deduced with base on workload, body mass, and age, which explained 89% of the VO2 variation, with SEE (standard error of the estimate) = 0.33 l.min-1. For the maximum load in the male group, another model with the same variables explained 71% of VO2 variation, with SEE = 0.40 l.min-1. For women, 93% of VO2 variation could be explained, with SEE = 0.17 l.min-1, both in submaximal and maximal exercise, with only one equation by use e of the same independent variables. CONCLUSION: The models derived in the present study proved to be accurate to predict submaximal and maximal VO2 in young Brazilian adults. (Arq Bras Cardiol. 2010; [online]. ahead print, PP.0-0)