We evaluate a modified version of the Park wake model against power data from a west-east row in the middle of the Horns Rev I offshore wind farm. The evaluation is performed on data classified in four different atmospheric stability conditions, for a narrow wind speed range, and a wide range of westerly wind directions observed at the wind farm. Simulations (post-processed to partly account for the wind direction uncertainty) and observations show good agreement for all stability classes, being the simulations using a stability-dependent wake decay coefficient closer to the data for the last turbines on the row and those using the WAsP recommended value closer to the data for the first turbines. It is generally seen that under stable and unstable atmospheric conditions the power deficits are the highest and lowest, respectively, but the wind conditions under both stability regimes are different. The ensemble average of the simulations does not approach the limits of the infinite wind farm under any stability condition as such averages account for directions misaligned with the row.