See supplement for reviewer comments.

This manuscript presents a comprehensive evaluation of the MOM6-NEP10k model's performance over the Bering Sea, along with sensitivity tests on turbulent mixing parameterizations. The work is rigorous, well-motivated, and makes valuable contributions to regional modeling, particularly in high-latitude domains where ocean–ice interactions and cold pool dynamics are essential for ecological forecasting.

The paper is timely and technically sound. However, the manuscript could be strengthened through greater clarity in presentation, more quantitative analysis in some sections, and deeper contextualization of the results, especially with respect to model biases and sensitivity test implications.

Line 20. Clarify "retreat earlier (later) in cold (warm) years"—this may seem counterintuitive. Explain mechanism briefly.

Line 56. “northeasterly winds in winter” → add citation or mean wind speed for reference.\

Line 86. Abbreviate NOAA

Line 94. Please provide citation that reports the excessive shear-driven vertical mixing found in MOM6-NEP default configuration. 

Line 95. The physical justification for the particular choices of scaling factors could be expanded. What guided the specific values tested? Clarify whether these fall within observed oceanographic ranges or are purely numerical tuning.

Lines 118-119. It could be confusing to the readers to mention the boundary that connects to the Pacific as "western" when most of it appears to be a "southern" boundary. 

Line 125. GLORYS12 is said to perform well in coastal areas—consider supporting with performance metrics from the cited studies.

Line 140. Can you mention the reasons/justifications for using a 3-hourly product vs 1-hourly product of ECMWF for atmospheric forcing?

Lines 305–360: “...model tends to underestimate sea-ice cover...melts all ice at the same rate...”

This is an important insight into model limitation. I'd suggest incorporating references to ongoing work or methods that include age-dependent melt rates. Could this be easily incorporated in future iterations?

Line 310: “mean ± std dev of areal ice extent” → good summary; you might also include coefficient of variation to show relative interannual variability.

Lines 365–400: “...agreement in summer was poorer...SST is highly sensitive...”

Consider plotting correlation maps of SST biases against ice retreat dates to support the hypothesis. A brief regression analysis could further validate the sensitivity relationship.

Lines 405–415: “MLD biases in winter increase significantly…”

This is a critical point. MLD overestimation in winter could lead to errors in nutrient entrainment and ecosystem models. Include a brief discussion on whether this is driven by JRA55 biases or turbulent mixing choices. 

Lines 440–460. It would help to add a histogram of thermocline depth differences (model – obs) across seasons. Also clarify the threshold logic—what happens when vertical gradients are weak or noisy?

Lines 510–530. Consider adding a short discussion of potential ecological implications (e.g., fish spawning habitat or match/mismatch hypothesis) to emphasize why this matters.