Huang et al. presents a comparison of 8 different air quality models/mechanisms in how these models produce secondary organic aerosol (SOA). They run these models offline, e.g., using a box model, so that same initial conditions are used and initial results can be compared. They find that each model/mechanism lead to very different yields of the similarly tracked species (aromatics, biogenics). Though this could be a good reference paper for modeling community and insight into how to compare and/or improve models, the paper currently as presented is not ready to be published in ACP, for the following reasons:



1) The first 16 pages of the manuscript currently reads more as a review of SOA modeling schemes than a research article. Though this compilation of models is useful for an easily digestable reference,  I recommend that the authors and/or editor determine if this paper should be a review, a research article, or a measurement report, as I will discuss next.

2) The results and discussion, currently as written, read more as a measurement report, meaning that the values on the figures are directly discussed without really place the results from these figures into bigger context. This bigger context includes directly comparing the results from one model to another, which is sometimes done but gets lost. Also, placing the results into the context of prior research is currently not done, leading the the results/discussion reading as a measurement report. To move the science beyond measurement report, if the authors want this to be a research article, I would recommend more in depth analysis (e.g., plotting the SOA yields as ratios to a model to demonstrate how much more SOA is being produced or plotting the average and spread of SOA yields across the models and what does that mean for total SOA produced for a typical urban environment and typical biogenic environment for uncertainty). Another demonstration of this is that figures are rarely referenced, esp. Fig 6, and again generally reads more as a review in that this is how model A performs, this is model B performs, etc. How much in how the model performs is surprising? E.g., if the authors compiled results of model comparisons or different model performances in a review type method, are the differences in yields and aging surprising?

3) Authors state that they do not want to say use model X after this evaluation; however, from the results, it would feel like at least two models need to be used in order to better demonstrate all possible answers and uncertainty, which would be resource heavy. Could any evaluation against published chamber studies be done to provide better guidance of at least potential biases of one model vs another?

4) The largest concern is the title does not reflect what is in the paper. After reading this, there was no evident discussion about the policy implications of these 8 different models/mechanisms. If this is important aspect the authors want to address, which I would support, more in-depth analysis, as suggested in comment 2), should be done. E.g., if a typical urban area starting with a mixture of aromatics, biogenics, IVOCs, etc., is modeled, how much total SOA is produced with the different metrics the authors discuss? How do these differences imply differences in policy strategy, such as explicit emission control (assuming most of the precursors are coming from heating or transportation or another source) vs more widespread emission control (transportation plus solvents plus . . .). What could the economic and/or public health impact be using one model that has lower total SOA vs another that has too high SOA? E.g., if the lowest performing model underpredicts the SOA for an urban area, what does that mean for the policy implications that were pursued due to using that model? If the model overpredicted, is that an economical concern?