Geo_point vs geo_shape differences

Hi All,
I am currently evaluating the migration from Elasticsearch to Opensearch.

Currently we use geo_shape, but the geo features we index are all points.
This means we could also use geo_point.

Many years ago, we have decided to use geo_shape as geo_point was having some issues when using aggregations.
See this post:

Now, i remeber that something changed for geo_point at some version of Elasticsearch.

So i guess my question is:
Is geo_point still stored just as two numbers on the index or is it stored as terms?

It seems that there is a difference in storage

green  open   track_202411                      ttAOVMiSTCWdTXwR5Gi-mw   6   0    1609882            0    391.2mb        391.2mb
green  open   geo_track_202411                  9sAoPRUjQsOJQC6f1ueacA   6   0    1609882            0    304.1mb        304.1mb

As you can see (after _forcemerge) we have 304 MB for geo_point and 391MB for geo_shape for the same number of documents (same documents actually) which is 25% less storage.

I would love to save some disk space, but if geo_point is bahaving in the same old way, of loading all documents in memory just to check if they are inside a polygon, this will be a hard no for the geo_point.

Thanks in advance
/Georgi