I have thoroughly enjoyed reading this fascinating work and am delighted to see that the IAP-CAS model has achieved a good prediction skill for the MJO, with a 24-day skill in the 20-year hindcast. It is important to investigate the predicted amplitude and propagation of MJOs, as highlighted by Chidong Zhang in the 2024 EGU meeting (personal communication)： The stable eastward propagation, with an average speed of about 5 m/s, is the most important feature of the MJO for sub-seasonal prediction. Additionally, the physical explanation presented for the prediction bias in terms of enhanced MJO amplitude and faster eastward propagation is also reasonable. This paper is well-organized, and the interesting findings justify its publication in Geoscientific Model Development. I have some suggestions and comments for the authors to further improve their work.

It appears that the IAP-CAS model has already been involved in the S2S project, and we can download the hindcast data from the S2S project. Is the model used in this work the same version as the one from the S2S project? Why not create a figure comparing the prediction skill among all S2S models? Some studies have conducted such comparisons, and it is necessary to present the average and best skill among the current S2S models.

 Was the prediction skill of 24 days calculated for the annual MJO or for the boreal winter MJO? It is important to clearly state whether the main conclusions are for the annual mean or for the boreal winter. (Sometimes you show the results for the annual mean, while some figures were drawn for the boreal winter), as S2S models exhibit a significant annual cycle in the prediction of MJO. Similar checks are also necessary for other presentations. For example, in Lines 168-170, you only have 16 ensembles since 2019, while in Fig. 3, you also presented 16 ensembles for the long period of 1999-2018.    

Line 42: The impact of MJO on sub-subseasonal prediction of each sub-monsoon precipitation has been well discussed (Liu et al., 2022), and should be referenced.

Liu, F., Wang, B., Ouyang, Y. et al. Intraseasonal variability of global land monsoon precipitation and its recent trend. npj Clim Atmos Sci 5, 30 (2022). https://doi.org/10.1038/s41612-022-00253-7. 

Lines 117-119: There were many phenomena that affect the MJO propagation. I suggest deleting this statement as it is not directly related to this work.

 Lines 145: I cannot follow why you use the 10-day forecast nudging from GFS forecast. Should we attribute the good prediction skill of 24 days to IAP-CAS or GFS? Lines 245: Was this underdispersive due to weak initial perturbation of the time-lag method? 

Lines 306-310: You can calculate the phase speed in this Hovmöller diagram directly. I have a different explanation for the phase speed difference. In Fig. 6, the predicted zonal scale of the MJO, represented by the easterly wind anomalies to the east of the MJO convective center, covers a larger region than observed, which is more obvious for the slow-propagating mode. The moist central Pacific in IAP-CAS overestimates the zonal scale of the MJO, which will increase the eastward propagation speed of the MJO, since the phase speed is inversely proportional to the wave number, as shown in previous work (Wang et al. 2019Sci. Adv. Diversity of MJO). The increased MSE tendency to the east of the MJO can explain the increased amplitude of the MJO, rather than the propagation speed. Let's make an assumption: for the same speed, the stronger MJO also has a larger MSE tendency to the east than the weaker MJO.

Please see my review in the supplement. I think the paper is solid and the analysis is interesting, although the system itself isn't particularly distinguished from S2S systems.