The paper exposes some recent new trends in modelling jets-in-crossflow with relevance to film-cooling of turbine blades. The aim is to compare two classes of turbulence models with respect to their predictive performance in reproducing flow physics. The study focuses on anisotropic eddyviscosity/diffusivity models and explicit algebraic stress models, up to cubic fragments of strain and vorticity tensors. The first class of models are DNS-based two-layer approaches transcending the conventional k–ε model by means of a non-isotropic representation of the turbulent transport coefficients; this is employed in connection with a near-wall one-equation model resolving the semi-viscous sublayer. The aspects of this new strategy are based on known DNS statistics of channel flows and boundary layers. The other class of models are quadratic and cubic explicit algebraic stress formulations rigorously derived from second-moment closures. The stress-strain relations are solved in the context of a two-layer strategy resolving the near-wall region by means of a non-linear one-equation model; the outer core flow is treated by use of the two-equation model. The models are tested for the film cooling of a flat plate, and are then extended to film cooling of a symmetrical turbine blade by a row of laterally injected jets. Comparison of the calculated and measured wall-temperature distributions shows that only the anisotropic eddy viscosity/diffusivity model can correctly predict the spanwise spreading of the temperature field and reduces the strength of the secondary vortices. The non-linear algebraic stress models were of a mixed quality in film cooling calculations.