A new subgrid-scale stress model based on the square of the velocity gradient tensor is proposed for Large Eddy Simulations (LES) in complex geometries. This model accounts for both the strain and rotation rate of the smallest resolved turbulent fluctuations. It recovers the proper $ y^3 $ near-wall scaling for eddy viscosity without requiring a dynamic procedure. The model is shown to handle transition in periodic turbulent pipe flow. The model is based on the square of the velocity gradient tensor, which is invariant to coordinate translations or rotations. It is a function of both strain and rotation rates and naturally goes to zero at the wall, eliminating the need for damping functions or dynamic procedures. The resulting Wall-Adapting Local Eddy-viscosity (WALE) model reproduces the proper wall scaling $ \nu_t = O(y^3) $ and is well-suited for LES in complex geometries with structured or unstructured methods. It is sensitive to both the strain and rotation rate of small turbulent structures. The WALE model is tested in isotropic turbulence and turbulent pipe flow. In isotropic turbulence, the model reproduces the experimental spectra with $ C_w \approx 0.5 $. In turbulent pipe flow, the model handles the transition to turbulence and reproduces the logarithmic law for the mean velocity profile. The WALE model produces a more accurate representation of the turbulent flow compared to the classical Smagorinsky model, especially in the near-wall region. It also shows better performance in capturing small-scale turbulent structures near the wall. The WALE model is invariant to coordinate translations or rotations and only requires local information, making it suitable for complex geometries. It is numerically well-conditioned, with the eddy viscosity neither negative nor infinite. The model is efficient and practical, as it does not require filtering at various scales. It has been successfully applied to freely decaying isotropic turbulence and turbulent pipe flow using a hybrid mesh. The model shows good agreement with experimental data and is well-suited for LES in complex geometries.A new subgrid-scale stress model based on the square of the velocity gradient tensor is proposed for Large Eddy Simulations (LES) in complex geometries. This model accounts for both the strain and rotation rate of the smallest resolved turbulent fluctuations. It recovers the proper $ y^3 $ near-wall scaling for eddy viscosity without requiring a dynamic procedure. The model is shown to handle transition in periodic turbulent pipe flow. The model is based on the square of the velocity gradient tensor, which is invariant to coordinate translations or rotations. It is a function of both strain and rotation rates and naturally goes to zero at the wall, eliminating the need for damping functions or dynamic procedures. The resulting Wall-Adapting Local Eddy-viscosity (WALE) model reproduces the proper wall scaling $ \nu_t = O(y^3) $ and is well-suited for LES in complex geometries with structured or unstructured methods. It is sensitive to both the strain and rotation rate of small turbulent structures. The WALE model is tested in isotropic turbulence and turbulent pipe flow. In isotropic turbulence, the model reproduces the experimental spectra with $ C_w \approx 0.5 $. In turbulent pipe flow, the model handles the transition to turbulence and reproduces the logarithmic law for the mean velocity profile. The WALE model produces a more accurate representation of the turbulent flow compared to the classical Smagorinsky model, especially in the near-wall region. It also shows better performance in capturing small-scale turbulent structures near the wall. The WALE model is invariant to coordinate translations or rotations and only requires local information, making it suitable for complex geometries. It is numerically well-conditioned, with the eddy viscosity neither negative nor infinite. The model is efficient and practical, as it does not require filtering at various scales. It has been successfully applied to freely decaying isotropic turbulence and turbulent pipe flow using a hybrid mesh. The model shows good agreement with experimental data and is well-suited for LES in complex geometries.