Please see supplemental file.

This paper presents a comparison of various numerical modeling tools while modeling wake steering operation for the IEA 22MW reference wind turbine. The paper is clearly written and presents useful results, and I believe the paper is publishable subject to a handful of comments.

My main comment is on the LES and whether it should be treated as the "truth" for comparison. I am not and expert on LES so I can't comment on the choice of simulation setup, although I have no reason to doubt that the authors have chosen a reasonable setup. However, I am interested in how closely the LES approximates true, atmospheric flows; and whether we should be treating the LES as the "truth" for comparison. For instance, my impression is that LES often contains lower wind direction variability than is usually observed in the atmosphere, and wind direction variability often strongly affects apparent wake location/expansion. Throughout the paper, the authors sometimes make statements comparing other models to LES, for instance " ... gives the closest results to LES", which I agree with. However, at other times, the comparison to LES is unstated/implied, such as "RANS underpredicts lateral deflections." My concern is that the implication is not strong enough, and some readers may believe that (in this case), RANS is "wrong", since it differs from LES.

I think the authors could take a few steps to remedy this. First, I would be interested for the authors to add a paragraph explaining how well the LES they used approximates observations of the atmosphere: what aspects of atmospheric flows is the LES capturing, and what aspects is it missing? Second, that any "poor performance" from other models is _only_ in comparison to LES should be made clear throughout (not necessarily in every sentence, but just so that there is no ambiguity for the readers). Third, can the authors quantify the uncertainty in the LES so that readers know, when there is a small difference between the LES and another model, if that difference is meaningful?

I have a few smaller comments and a handful of minor typo corrections etc, as listed below.

- Is the DWM driven by flow after filtering at the cutoff frequency f_c? Or is the filter somehow built into the DWM model? Either way, how is the filtering applied to remove high-frequency content from the turbulence that drives wake meandering? What is the form of the filter?

- Line 389: "It is clear that the LES and DWM signals are correlated". This is not really obvious to me looking at Figure 7, and besides is not very precise. Can the authors quantify how much correlation there is between the center positions (lateral and vertical) for DWM and LES?

- How is secondary steering defined? That is, how is the second turbine's wake isolated from the first turbine's wake?

- What computational resources (computational hardware and computer wall time) are needed for each of the models? A table comparing the models in terms of computational complexity would be interesting for readers.

- The term "benchmark" not really defined. As I understand it after reading the paper, this is a "benchmark test" in the sense that multiple models are compared; however, it is not a "benchmark" that others can compare to, as I was originally expecting based on the title. Consider removing the term benchmark or clearly defining what you mean by "benchmark".

- Add outlines to Tables 1 & 2, they are a bit difficult to understand without outlines. 

- What range of rotor-average wind directions are observed over the simulation?

- Am I understanding correctly that using LES for the background flows for the DWM is ok because it only needs to be run once (whereas the DWM model itself could be run many times to compare controllers)?

- Is the inflow WS, TI used for PyWake the average over the hour from the LES?

- Why is the Hill Vortex deflection model used in DWM but JimenezJimenez used in PyWake? Is that just to do with what is implemented?

- "Half a degree of wind direction offset corresponds to roughly three degrees of yaw steering"---what do the authors mean by this? Do they have analysis or citations to support this?

- Fig 2. Is the dramatic change in deficit in the near wake when steering in the DWM expected?

- Fig. 6: Why do the colors in the legend not really match the bars? Also, there appear to be four "textures" defined, but in the figure I only see the "HAWC2 ghost turbine" and "velocity + power curve" cases in the plot. (or possibly the "HAWC2 turbine" does appear for DWM only at x/D = 0)?

- 397 "max out at" is a bit informal---consider revising.