Wind turbines installed in a wind farm are typically subject to increased turbulence because they are in the wake of upstream wind turbines that generate additional turbulence. Accurate prediction of turbulence in wind farms is critical as it is proportional to wind turbine fatigue loads, power losses and prediction of wind farm lifetime. IEC Standard 61400-1 suggests the use of the semi-empirical turbulence model called Frandsen model, which was originally proposed in 1999. Since the development of the Frandsen model, the size of wind turbines and wind farms has increased significantly. Therefore, this work aims to determine the accuracy of two versions of the Frandsen model: Standard and Modified, when applied to large offshore wind farms experiencing a combination of atmospheric stability conditions. This is done by comparing the estimated wind farm turbulence under specific atmospheric stability conditions with measurements from Westermost Rough and Horns Rev 2. It was found that the atmospheric stability distribution and the distance of the upstream wake inducing wind turbine at the offshore site plays a significant role in the accuracy of the estimated turbulence from the Frandsen models. The estimated turbulence intensity from the Standard and Modified Frandsen models was found to be under-predicted with respect to the measured turbulence for all atmospheric stability conditions at both wind farms. For wind directions with wake flow for upstream wind turbines more than 10 rotor diameters away, the Modified Frandsen model showed better prediction of turbulence intensity compared to the Standard Frandsen model. On the other hand, for wind directions with wake flow for upstream wind turbines less than 10 rotor diameters away, the Standard Frandsen model showed better turbulence intensity estimation compared to the Modified Frandsen model. Among all atmospheric stability conditions, the turbulence intensity estimate was closest to the measured data for unstable conditions. It is suggested that this fact can be attributed to the presence of a significant number of unstable conditions in the offshore wind farms used for the design of the semi-empirical Frandsen model.