Multifield models of inflation predict an inequality between the amplitude τNL of the collapsed limit of the four-point correlator of the primordial curvature perturbation and the amplitude fNL of the squeezed limit of its three-point correlator. While a convincing detection of non-Gaussianity through the squeezed limit of the three-point correlator would rule out all single-field models, a robust confirmation or disproval of the inequality between τNL and fNL would provide crucial information about the validity of multifield models of inflation. In this paper, we discuss to which extent future measurements of the scale dependence of halo bias can test multifield inflationary scenarios. The strong degeneracy between the effect of a non-vanishing fNL and τNL on halo bias can be broken by considering multiple tracer populations of the same surveyed volume. If haloes down to 1013 M⊙ h−1 are resolved in a survey of volume 25 (Gpc3 h−3), then testing multifield models of inflation at the 3σ level would require, for instance, a detection of τNL at the level of τNL ∼ 105 given a measurement of a local bispectrum with amplitude fNL ∼ 10. However, we find that disproving multifield models of inflation with measurements of the non-Gaussian bias only will be very challenging, unless |fNL| ≳ 80 and one can achieve a halo mass resolution of ∼1010 M⊙ h−1