The manuscript "ISMIP6-based Antarctic Projections to 2100: simulations with the BISICLES ice sheet model" by O'Neill et al. describes and discusses an ensemble of projections of Antarctic ice sheet change to 2100 simulated with the BISICLES ice sheet model. Though these simulations are essentially done as a contribution to ISMIP6, they were not included in the original ISMIP6 results release, though they did contribute to the later analysis of Edwards et al. (2021) and played a role in informing the most recent IPCC sea level projections. Thus, it is important for these results to be described in a peer-reviewed publication given their contribution to widely used sea level projections.

Overall, this is pretty straightforward paper describing model results, following the established ISMIP6 protocol. While I did find it to be a bit dense in places, in general I think the detail was appropriate given the nature of the results. I can't quibble much with the modeling choices since most of them are dictated by ISMIP protocol, and so configuring the model and simulations in order to be able to compare them with the rest of the ISMIP ensemble is sensible. Below I do point out one aspect of the science/presentation which I think needs to be improved upon.

I do think much of the "discussion" section felt like just further description of results. In particular, I would suggest that section 4.1-4.4 be moved into section 3 since they are mainly a description of the results without much discussion or comparison to other studies. I also think these discussions of the results should probably be condensed by maybe 20-30% for the sake of readability.

Perhaps my main concern with the science in this paper is how the control run is treated and discussed. I understand that it is standard ISMIP6 practice to subtract the control run from all results such that the resulting numbers represent "sensitivities" of the ice sheet to future emissions forcing. However, this procedure is then somewhat at odds with presenting the results as true projections of future sea level rise, as they are in this paper. While here in this paper and in other ISMIP6 publications the control is often discussed as representing model "drift", it does lump together many potential real sources of ice sheet change including the transient evolution of the initialized ice sheet state, which is out of equilibrium. The paper says as much around lines 220-221 where it says "Whilst subtracting the control can account for model drift, it may also in this instance be removing the sea level signal from ASE’s long timescale to retreat initiated before 2015". I think this is quite important because the control run here simulates a non-trivial contribution to sea level rise (6 cm), comparable at first order to the forced changes simulated in the non-control simulations. Thus, when the paper says (e.g.,) that so-and-so simulation represents a "sea level fall", this isn't accurate. Rather, such simulations represents less sea level rise than in the control simulation, but the raw simulation is in fact projecting sea level rise (since even the most "sea level fall" is 5.3 cm, less than the SLR in the control). This can be quite confusing for a reader who is looking to this paper simply for sea level projections. My suggestion would thus be to revise the language throughout the text to discuss the projections as being relative to the sea level rise simulated in the control (i.e., not a "sea level fall" but "less sea level rise than in the control", and not "sea level rise" but "more sea level rise than in the control"). You say something like this for one part of the analysis (line 201), but it applies to all the analyses presented in this paper. Alternatively, you can just not subtract the control run in the plotted results as presented, while still discussing the control run at length. I understand that this is a departure from ISMIP6, but given that we are already moving on to ISMIP7 as a community, this paper could point to a better way to think about considering the control run.

Beyond this issue, my other suggestions are minor and listed here:

L3: ice sheet

L4: 21st century, quantifying uncertainty...

L7: to the Antarctic projections

L13: delete "tested"

L15: were the third largest contributor to global mean sea level (GMSL) change from 1901 to 2018, behind...

L16: delete "dominating the...between 1901 and 2018"

L22: dominated by the dynamic glacier response to warm

L25: ice sheet may be vulnerable

L28: warming, ice loss from marine basins...

L30: Ritz et al. 2015), amplifying uncertainty...

L31: increased Antarctic snowfall accumulation rate has...

L36: To better project

L53: further ensemble of simulations using

L62: the ISMIP6 design for

L78: volume, ice sheet model solving the L1L2 flow approximation with adaptive mesh

L90: It would make more sense to say that you set the rate factor in effective viscosity purely based on temperature and then you also invert for damage given the A(T) field. (If I understand properly). Not sure if this is different than just inverting for A, but perhaps I don't understand.

Section 2.2: how is basal melt treated at/across the grounding line?

L112-114: fix some citation rendering

L120: Am I understanding correctly that this is the ISMIP6 protocol for calibrating gamma_0? If so, perhaps than say that more clearly. If it isn't, it would be good to know why these older datasets of basal melt were used, not the more accurate and comprehensive recent basal melt data (e.g. Paolo et al. 2015, Adusumilli et al. 2021).

L142: I'm quite confused about why basal melting is applied in this way for the control run? Is it time dependent? Does it vary as the model evolves or is it prescribed at the beginning and then held constant. I'm not sure how I understand the sense in which this is a control. Don't other ISMIP6 models just apply a constant-in-time basal melt forcing for the control?

L154: citation misrendered

Section 3.1: the paper mainly discussed how the forced simulations compare to other ISMIP6 models, but it would be useful to know how the control simulation compares to them as well

L160: is the reason for slow down at major ice shelves the lack of calving in this model?

Figure 3: colorbars missing variables and units

L180: is it possible that the increase in floating area is causing an increase in buttressing. This is an artifact of models that fix the calving front, as discussed in Haseloff and Sergienko 2018, and may have considerable upstream effects on marine ice sheet stability

L215: sea level rise or fall?

L217: sea level rise

L220: sea level rise

L234: model parameter

Figure 7: to me it would make mroe sense to have the AP panel with the same y-axis as the other panels to emphasize the very different scale of contribution, but can understand if the authors would prefer to keep it this way for legibility

L270: I am confused by this sentence since the choice of gamma_0 is independent of the choice of GCM. I can see how the result is dependent both on GCM and gamma_0, but not how one is dependent on the other. Perhaps more explanation is needed.

L279: explain why only MeanAnt was sampled

L331: perhaps a simpler way to put this is that the ice sheet response to ice shelf collapse is independent on climate forcing (i.e. it is purely ice dynamical)

L364: long-term

L384: It is known that models with friction interpolated across the grounding line/zone are more sensitive and tend to have larger response to forcing than models with more conventional schemes (Tsai et al. 2015). It seems like that is probably playing a role here.

L390: this is where it would be good to know how basal melt at the grounding line is treated in BISICLES, since this has a big influence on the model sensitivity as Seroussi and Morligem showed.

L395: technically, the sliding scheme in BISICLES is closer to Tsai et al. 2015 than a purely Weertman sliding law, which may be a point of difference with ISSM.

L410: this is because there is no floating ice left on which basal melt can be applied - useful to point out

L424: projections should be informed

L432: is the model computationally expensive? would be useful to have context for why more simulations were not done for this study

L456: increasing snowfall accumulation

L475: I would hope that the statement "Data is available on request, and will be publicly available in due course" is merely a placeholder for the pre-print, since I'm not sure it is useful for a paper publishing important results contributing to widely used sea level projections. My suggestion would be to make these data available before the paper is published.

L498: X?

Please see attached pdf.