Give that these hexbins are made to look like regular hexagons on a merkator projection, there is no easy answer to this question.
The only real metric you can get is the longitude extend of one hexagon: it is equal to the longitude extent of your points divided by the number of horizontal hexagons:
(lon_max - lon_min) / nx_hexagon. This gives you the horizontal size of a hexagon in degrees.
Then to convert it to meters, you would need to use the Earth radius and the local latitude. This can be approximated if your data does not span on too large a scale in terms of latitude. If you are in conditions where the latitude extent is small (~1 degree max, especially further away from the equator), you could get the horizontal size of the hexagon with the following formula:
earth_radius = 6.371e6 # meters
# Horizontal size in meters
size_x = (lon_max - lon_min) / nx_hexagon * np.pi / 180 * earth_radius * np.cos(lat_mean)