Future digital subscriber line (DSL) systems are supposed to achieve higher data rates by using interference cancellation techniques. Since far-end crosstalk (FEXT) is the major impairment in DSL systems based on frequency division duplexing, mitigation of this interference can increase the signal- to-interference-and-noise ratio, and therefore significantly boost the data rate. For simulation and evaluation of FEXT cancellation algorithms, more accurate modelling of the multiple-input- multiple-output (MIMO) crosstalk channels is needed. Until now, DSL standards usually rely on the 99% worst case modelling of crosstalk power-sum for the lines in a binder. This paper proposes a parametric stochastic FEXT model based on the sum-of-sinusoids approach for the pair-to-pair crosstalk among multiple lines in a binder, which shows a more realistic behaviour than the worst case models.