We consider a class of linear regression model with extreme distribution noise. We show by a mean of point process technique that the asymptotic distribution of the maximum is the same as the one of the max of the noise process, under specific conditions.