In this work models of sh011 term are considered to foretell the inflation of transables and non transables goods in Colombia. These models did not exist in the Central bank before 2004 and are very useful for the decision making of monetary policy. Also the benefits are evaluated, in capacity and analysis terms prognosis, to use methods that capture the possible nonlinearity of the curve of Phillips in the Colombian data. Although different reasons exist that justify a relation nonlinear of sh011 term between product and inflation, each one of them suggests a different form for the curve. Therefore, the flexible square minimums are used anifieial neuronal networks (ANN) and (FLS), procedures that have the great advantage of which they beforehand do not impose any functional form that can Slant the results. Once the estimation is made of the models of inflation of transables and of non transables, the prognoses of these two nonlinear models with those of two linear estimations are compared, the impulse functions are analyzed answer of each one of the models and in addition 3 test of nonlinearity is made. One is that the curve of Phillips in Colombia could be nonlinear and therefore turns out pertinth to consider models nonlinear for his estimation. Finally, with these models it is tried to explain the process of desinflación that has lived the Colombian economy in the last years as much in the inflation on transables, like in the one of non transables.