Swirling flows are very dominant in applied technical problems, especially in IC engines, and their prediction requires rather sophisticated modeling. An adaptive low-pass filtering procedure for the modeled turbulent length and time scales is derived and applied to Menter' original k - ω SST turbulence model. The modeled length and time scales are compared to what can potentially be resolved by the computational grid and time step. If the modeled scales are larger than the resolvable scales, the resolvable scales will replace the modeled scales in the formulation of the eddy viscosity; therefore, the filtering technique helps the turbulence model to adapt in accordance with the mesh resolution and the scales to capture. The novel turbulence model presented in this work will be called Dynamic Length Scale Resolution Model (DLRM), because of its capability to dynamically adapt its behavior according to the grid resolution and to consequently switch from modeling to resolving the turbulent length scales. Validation has been carried out both on a strongly swirling flow through a sudden expansion and on a simple IC engine geometry with one axial central valve; the model seems able to capture unsteady effects and to produce accurate time-averaged results (especially if compared to its standard RANS formulation) and looks particularly suitable when used with grids where turbulence would not be sufficiently resolved for an accurate LES.