The paper is an essential contribution to modeling GMFs, especially snow avalanches. During the last couple of decades, most work has been focused on physical models, and most work has been published as proprietary software with high fees to use. These models are still widely used today, but the proprietary software limits scientific development. The Flow-Py tool is a new and improved model built upon existing literature/models. Finally, seeing an open-source model published will be a significant step forward where other researchers can implement and improve the model.

Both the paper and the code are well written. I'm able to run the code without any problems. One note: several times, it's stated that the code is computationally easy to run. My impression after running the code is that it is relatively computationally heavy to run. The authors do not compare the processing time to similar physical or data-based models, but this is subjective. I'm not sure if, i.e., the RAMMS software for the same 100 km2 extent will take more than the roughly 4 hours mentioned to run the AOI.

How the flux controls the spreading and partly Z is well explained in Figures 3 and 4. In Figure 5, there is an example where the GMF has to overcome a dam in the slope. It would be nice to see an example where the kinetic energy isn't high enough to overcome the dam, and the total runout doesn't reach the given threshold travel angle.  

Line 35-36: What is the order of citations? By year (old to new)? Alphabetical? It's not consistent throughout the paper.

Minor spelling details:

Line 226: "cells"

Line 227: Fix degree symbol

Line 285: No capital letter in "output"

Page 24, footnotes: "thee"

This paper is an important advancement in the study of GMF and snow avalanches, especially for studies at larger scales and in data space areas.   Specifically the work has significant potential to assist with further automation of regional avalanche terrain exposure scale mapping at larger spatial scales. I'm glad to see open-source, object-orientated code allowing for further development and adaptation.

The paper was well written, at the validation methods applied were appropriate.  I agree with other commentators that it would also have been interesting to see a comparison of processing times for the same study path using different models.  It would have been interesting to see a comparison of the raster output of the model against other GMF approaches, such as TauDEM.

The code was well documented and there was sufficient guidance to configure the parametrization.  I was able to run the code run on both Linux and Windows machines and replicate the sample studies.  In practice, I found it was most efficient to conduct simulations over larger spatial scales using an AWS cloud computing implementation to fully benefit from the parallel processing.

Technical corrections:

ln 15 - topograhies -> topographies

ln 86 - modeling -> modelling

ln 399 the these  -> these