The evaluation of carcass composition makes possible to assign a value for the carcass. It is possible to predict carcass composition by measurements taken on the carcass: weights, carcass yield, fatness, conformation and other non-invasive methods. The use of these measurements offers an advantage over joint dissection, as they are faster and easier to make and do not involve any loss of carcass tissue. The aim of this work is to develop equations to predict the tissue composition of guinea pig carcasses, by using weight, yield and linear measurements. The data were obtained after a practical and normalized method to measure the guinea pig carcass, based on anatomical guidelines. The carcass measurements of 40 South American Guinea pigs, males and females of 3 and 12 mo age (fattening and discarded animals, respectively) were used for multiple regression analysis; which was implemented from the correlation matrix of dependent and independent variables and the significance test. The independent variables or predictors used were live weight at slaughter (LWS), hot carcass weight (HCW), perirenal and pelvic fat (PPF), external carcass length (ECL), and thorax circumference (ThC). The accuracy of the predictions was evaluated with root mean square error (RMSE) and coefficient of determination (R²). The prediction equations for carcass composition in grams were more accurate than those related to composition proportion. In the prediction of some variables in grams such as total muscle (TM), skin (S) and muscle + freeze-thawing loss (MpFL) were obtained values of R² = 95%, for total bone (TB) was obtained an R² = 90%, and for subcutaneous fat (SF) and total fat (TF) were obtained R² = 78% and 76%, respectively. Lean (or muscle) and fatness are both the most important commercial components of a carcass. In this case, PPF was a good predictor for guinea pig carcass fatness, whereas HCW and ThC were good predictors of lean content.