Is far as I can see, all models from the [Acoustic Toolbox][1] can be easily called from Matlab and Python.

My own preference is Bellhop, but have not used the latest version.

My experience with RAM is that wrong step/grid-size selection (related to frequency) can easily blow up computing time.

I would suggest to only use normal-mode or parabolic-equation based models for low frequencies (< 1kHz) and Bellhop otherwise. 


  [1]: https://oalib-acoustics.org/models-and-software/acoustics-toolbox/