Homeostasis model assessment of insulin resistance (HOMA-IR) is a method to measure insulin resistance. HOMA-IR cut-offs for identifying metabolic syndrome might vary across populations and body mass index (BMI) levels. We aimed to investigate HOMA-insulin resistance cut-offs that best discriminate individuals with insulin resistance and with metabolic syndrome for each BMI category in a large sample of adults without diabetes in the baseline of the Brazilian Longitudinal Study of Adult Health (ELSA-Brasil). Among the 12,313 participants with mean age of 51.2 (SD 8.9) years, the prevalence of metabolic syndrome was 34.6%, and 60.1% had overweight or obesity. The prevalence of metabolic syndrome among normal weight, overweight and obesity categories were, respectively, 13%, 43.2% and 60.7%. The point of maximum combined sensitivity and specificity of HOMA-IR to discriminate the metabolic syndrome was 2.35 in the whole sample, with increasing values at higher BMI categories. This investigation contributes to better understanding HOMA-IR values associated with insulin resistance and metabolic syndrome in a large Brazilian adult sample, and that use of cut-off points according to ROC curve may be the better strategy. It also suggests that different values might be appropriate across BMI categories.