Looks like it reduces initial cost of creating structures, etc for JSON data received and makes accessing the data easier.

The author does seem to say the library is more focused on reading the data rather than writing it.

Whether or not this is a benefit or not is up to you and your use case I guess.