Weiffenbach et al. assess the state of the Southern Ocean in 15 simulations of the mid-Pliocene performed as part of the Pliocene Model Intercomparison Project (PlioMIP2). 7 simulations with the full mid-Pliocene forcing (Eoi400) are also compared with experiments in which only the mid-Pliocene CO2 gas concentration was implemented (E400). This allows to separate the effects of the higher greenhouse gas concentration from the mid-Pliocene changes in orography. It is found that both the greenhouse gases and Pliocene orography contribute to the high southern latitudes climate change: i.e. warming, higher precipitation, stratification…  The manuscript is well written, and the results are interesting. As such I recommend publication in Climate of the Past after the comments below are taken into account.

1. Experimental design: At the moment the manuscript is targeted for readers who are familiar with the PlioMIP2 experiments. There is no detail in the methods regarding the experimental design used in these experiments. Readers are simply referred to Haywood et al. 2016. Please add some information in the methods regarding the experimental design used in the PlioMIP2 experiments so that a reader of this manuscript does not have to read Haywood et al., to fully understand the study.
2. Discussion: The discussion needs to be expanded and restructured.

While the discussion focuses on the relevance for future climate, there is a real paucity of references to other studies in the Discussion. There is no discussion of previous multi-model studies of the mid-Pliocene, previous multi-model studies focusing on high southern latitude climate, or previous studies discussing changes in SO stratification and the relation with the abyssal cell. For example, the paragraph starting L. 301 could be expanded to compare the results of the study with previous results.

The model-data comparison is limited and only shown in the SI. I understand there is limited data availability for this time period, but have the authors performed a literature search on the topic? Even though the overlap is small, the authors could look at Grant et al., 2023 (https://egusphere.copernicus.org/preprints/2023/egusphere-2023-108/) for marine sediment cores in the SW Pacific.

Finally, the order of the sub-sections in the Discussion should be thought through so that there is a better flow: i.e. The model-data comparison could come first, then discussion of results in the context of previous studies, relevance for future studies and caveats associated with SO biases.

3. Minor points:

• L. 200: high-latitude North Atlantic?
• L. 219: Relationship between changes in abyssal cell strength and OHT is interesting and consistent with previous studies (e.g. Menviel et al., 2015, EPSL).
• L. 301-302 There is not enough information to understand the relevance of the sentence (i.e. which scenarios? When?).
• L. 310-311: I don’t understand this sentence
• L. 319: “sea-ice cover”
• Figure 7: Is the cooling within the annual sea-ice zone due to the increased stratification?It does correspond to the area of large negative SSS anomalies. Is there reduced deep ocean convection in the Ross Sea sector in most models?
• Figure 8; The fit between precipitation and SSS anomalies is not particularly obvious. I am not sure P-E is a better fit. Should you look into changes in sea-ice melt/freezing, Antarctic runoff and eventually changes in ocean currents to understand the changes in SSS?
• Figure S11: Historical and not pre-industrial simulations should be compared with present-day observations.

Overall, I think this is good paper and should be published in climates of the past with little modification. The topic is appropriate for the journal and quite important in general. I have some suggestions that I think would make the manuscript appeal to a wider audience, which are mainly technical.

L35: you refer to a concentration pathway as being an emission scenario. This is confusing, as there is an esmrcp85 variant, but not a esmrcp45.

L97: justify why cosmos is treated differently

L112: What is the magnitude of the errors introduced by using monthly mean temperatures and salinities to compute the potential density and therefore stratification index?

Table 1: The first 4 models in this table are all variants of CCSM4. You should at least acknowledge this unbalanced ensemble design, and preferably comment on whether this could skew your findings (I suspect not).

Fig 1: At no point do you say in this figure caption what temporal averaging is being used. I presume this is annual mean, but showing a winter and/or summer sea ice edge is more conventional.

P9: I am very surprised that you are able to discuss AMOC in PlioMIP2 without citing Zhang et al’s 2021 paper called “Mid-Pliocene Atlantic Meridional Overturning Circulation simulated in PlioMIP2”. This work predates the Weiffenbach et al (2023) paper that is cited on the topic by 2 years! More disturbingly, 13 of the 16 authors of the present manuscript were also authors of the 2021 work. So it seems implausible that you (collectively) are unaware of it. Most importantly, if you (collectively) now have reservations about your (collectively) earlier work then you should be airing them publicly so that the whole community is aware of them.

Fig captions. Please do not start figure captions with uncommon acronyms. When I read any paper, I will first read the abstract and then look at the figures: only afterwards working my way through the main text, if I’ve found it interesting. I'm sure that I'm not the only person who approaches papers in a similar fashion. I recommend that you try to make your figure captions somewhat standalone. The current reliance on only PlioMIP2 terminology is pervasive. I was also confused by SMM, as even expanding it to “special model mean” does not help explain what the figure is.        

Fig 10: Consider adding the non-special models (boring models?) to this figure. They would not need to identifiable.

L290: Be wary of using “CMIP projections” when I think you mean the scenarioMIP projections. The ice sheet model intercomparison project is part of CMIP.

L294: I had forgotten what ~24m relates to, please remind the reader by providing more context.

L302: You mention CMIP5 studies here. Has no-one published anything on the topic relating to CMIP6?

L321: You are missing a closing bracket