The paper presents the new version of SPEEDY, its coupling to a three-layered thermodynamic ocean model, and some pace maker experiments. The paper is overall well written, but there a couple of points that should be addressed to make it suitable for publication.

Major comments

1. The idea of prognostic equations for sea ice thickness is interesting (mode 2), but I found the discussion of mode 1 (prescribed sea ice thickness) and mode 2 confusing. As I understand, mode 2 has not been tested. Given this and that you don't use mode 2 in the paper, it does not seem appropriate to introduce the prognostic computation of sea ice thickness in this paper.
2. In section 4 you begin with a simulation of SPEEDY-TOM3 in pacemaker mode. However, there was no discussion about the model's climate and how it compares to the prescribed SST runs described in section 3. Has coupling with TOM3 altered the model climatology? Also, to what extent does the model’s reproduction of global warming relate to the pacing of Indo-Pacific SST (and prescribed sea ice). The performance of the SPEEDY-TOM3 configuration without pacing the SST and prescribed sea ice should be described, as this would be a more natural control experiment (which prospective model users would need to perform).

Detailed comments.

L21 "observed statistics" -> "that observed"

L76-80, it is not entirely clear why this information is included in the introduction. Is it an important new feature of the model?

L110-111, why is there an interpolation to a finer horizontal grid?

L230, I assume that you must then add a corresponding term to the third layer (equal and opposite to the fluxes in the upper two layers) to account for the convective mixing also.

L261, Have you used a value of 1.5m for ice thickness in this paper? Please make this clear.

L322, Fluxes from AGCM simulations are known to be inconsistent with SST in regions where the SST variations are driven by turbulent fluxes, as in such AGCM experiments the SST are not influenced by the fluxes. Thus, it is not clear to me why you prefer to use an AGCM experiment for estimating fluxes.

L333-338, This could be written a little more clearly, by indicating the difference for the global average solar fluxes, and then writing "global average of the net surface heat flux" (rather than "net global averages").

In figures 4 and 5, it would be useful to also show differences maps. I can see differences between the ERA and SPEEDY patterns, but the text glosses over them.

It would be useful to show biases for rainfall shown in figure 7.

It is confusing to refer to "ens. 653" in the caption and the figure titles.

L498, it is not possible to understand the initial conditions for the TOM3 from this sentence.

L531-533, it would be useful to see a simulation without such changes. How does the model behave without prescribing changes in tropical SST?

L583-585, It should be explained that SPEEDY-TOM3 as described here has an imposed tropical SST warming. This may explain the large agreement with observations.

L616-617, Is the increased signal of the PNA in the North Atlantic Ocean an indication that O-A is important for this teleconnection? Or does it reflect some differences in the model climatology between the SST and TOM3 experiments?

L686, I think you mean "coupled and prescribed SST ensemble simulations". I am wondering whether the analysis from the models is computed using the ensemble means, and how this might affect the comparison with the observations (which is one realisation).

L688-689, this information is missing from the caption.

L698, do you mean "both coupled and uncoupled ensemble simulations"?

L709, I don't think that you need "natural" here.

L711, "runs of" -> "runs"

L769-770, There is a typo here as Tau_1 is not indicated.

L801, I don't follow why you say the ice mass cannot grow beyond an ice fraction. I think you mean the maximum rate occurs when fi = 0.5^0.5; It is also confusing that you discuss thickness above, but actually you use ice fraction in A6.

Review of the manuscript 'Multi-decadal pacemaker simulations with an intermediate-complexity climate model'

by Molteni et al.

The authors describe the major changes (new features) of a new version of the AGCM SPEEDY, and the formulation of a 3-layer thermodynamic ocean model (TOM3).

The most important change of the AGCM SPEEDY is in the interface between atmospheric and surface variables, and the

calculation of the surface fluxes. They present results on the climatology of the atmospheric model,

highlighting the influence of the modifications introduced in the AGCM SPEEDY, and shown how important features of interdecadal

and interannual variability are simulated in a pacemaker coupled ensemble of 70-year runs, where regions of the tropical

Indo-Pacific are constrained to follow the observed SST variability.

Major comments:

- the text is well written

- model seems to be a very useful tool

- Section 3.1: A comparison with the heat fluxes of the previous version of the model would be useful in the main text (similar to section 3.2).  Have the heat fluxes improved in the newer version due to the modifications?

- Section 4.1: Figure 10 a and b: I think it would be useful to have similiar Figures for the NH extratropical SSWT and extratropical land TAS anomaly. Wondering if 

the good agreement for the global means are dominated by the tropics - large parts of the tropical SSWT are prescribed.

- Section 4.3: Figure 14 and corresponding text: My impression is that there are substantial differences between the coupled model and ERA.

The coupled model patterns have a structure that reminds me of the NAO, while the ERA patterns remind me more of the East Atlantic pattern.

Therefore, the physical processes involved in the COWL-pattern might be different in the model and in the reanalysis data - in particular the importance of a dynamic ocean cannot be ruled out.

Minor comments:

- line 401 fig 7b => Fig. 7b

- line 686 Using data from both our coupled and ensembles => Using data from both our coupled and uncoupled ensembles?