This manuscript presents the development of a global dataset of column-averaged dry-air mole fraction of CO₂ (XCO₂) at high resolution (0.05°) using multi satellite products, and an improved deep learning model. They further evaluate new datasets using the measurements from the TCCON network. While the study shows promise, there are significant concerns regarding methodological transparency, and clarity of explicit demonstration of advantages over existing satellite products. Addressing these issues will greatly enhance the manuscript’s impact and originality.

1. The manuscript lacks in describing the methodological details of the improved deep learning model. The authors should clearly outline the specific innovations or modifications that lead to improved accuracy. Additionally, since the formation and distribution of XCO₂ are influenced by atmospheric transport processes across multiple vertical layers (and not solely by surface fluxes), it is important that the manuscript explains how these vertical transport processes are incorporated into the model. If these processes are not accounted for, this limitation should be explicitly acknowledged.
2. Although the authors present a new global XCO₂ product at 0.05° resolution, the distinct novelty and advantages over existing datasets remain unclear. The study should explicitly state how the analyzed information significantly differs from or improves upon existing satellite data. While validation against TCCON is good, the authors should explicitly compare these results with those from existing datasets to clearly demonstrate accuracy improvements.
3. The study lacks specificity in demonstrating how the new dataset quantitatively improves understanding relative to existing satellite data. Providing explicit examples or quantifiable differences would enhance the significance of this study.
4. The conclusion stating "promising advancement" is too broad. It should be explicitly clarified what specific policy, modeling, or scientific implications this advancement has, thus highlighting concrete applications or benefits.

This study reconstructed a global full-coverage XCO₂ product with a 0.05º spatial resolution using multi-component satellite data and an advanced deep learning method. However, the manuscript lacks innovation and sufficient detail in several aspects. My comments are listed below:

• The manuscript notes that the spatial resolution of current global full-coverage XCO₂ products is relatively coarse, ranging from approximately 0.25º to 2º (Line 128). However, global XCO₂ products with a 0.1º spatial resolution already exist (https://doi.org/10.1016/j.envint.2023.108057), indicating a need for more comprehensive literature review. Although this study improves the XCO₂ spatial resolution to 0.05º, its innovation and advantages compared to other datasets remain unclear. It is recommended to clearly articulate the study’s novelty and specific strengths.
• The model methodology section lacks essential explanations. The study employs the Attention-based Bidirectional Long Short-Term Memory (At-BiLSTM) model for global XCO₂ reconstruction, but it does not justify the choice of this model or clarify its advantages over traditional LSTM models. Additionally, the model’s interpretation remains unclear. It is recommended to provide a rationale for selecting At-BiLSTM and elucidate its specific benefits and interpretive framework.
• The discussion section requires further elaboration. It should comprehensively address the advantages of the model used and the resulting full-coverage XCO₂ product compared to other models and datasets. Additionally, the global spatial distribution characteristics of XCO₂ need more detailed discussion.
• In the conclusion or discussion section, please clearly specify the concrete data or scientific significance of the high-resolution XCO₂ Additionally, provide an outlook for future research, outlining key issues to address in global XCO₂ or CO₂ concentration reconstruction studies, such as critical challenges or priorities that should be focused on.
• The construction of the OCO dataset is unclear. For instance, it is not specified how grids containing both OCO-2 and OCO-3 data within the same time period were processed.
• The study utilized various satellite-derived variables, including land flux, anthropogenic flux, and climatic impacts, for global XCO₂ However, it is unclear whether these satellite data have gaps, particularly in high-latitude regions. If gaps exist, the study should specify how they were addressed.
• Line 243-244. “…spatial resolutions to 1 km resolution”. The ‘1 km resolution’ is inconsistent with the study’s focus on a 0.05º
• The manuscript contains several typo errors. For instance, in Line 303, “Figure 5. (a) Density scatterplots of sample-based…” includes an unnecessary ‘(a)’. In Line 314, the value ‘1.21’ should be corrected to ‘1.22 ppm’. These errors should be revised for accuracy and clarity.