Performance: Query large JSON Blob vs. actual graphql data

What query would be more efficient? The one that has to look up one field (Query A) and return a huge string or the one that has to look-up many very small fields (Query B) ?

Schema A:

type Foo {
  data: String!
}

Query A:

queryFoo {
   data 
}

Schema B:

type Foo {
   data: Data!
}

type Data {
   label: String!
   polygons: [Polygon!]!
}

enum PolygonType {
   A
   B
}

type Polygon {
   type: PolygonType!
   points: [Point2D!]!
}

type Point2D {
   x: Int!
   y: Int!
}

Query B:

queryFoo {
   data {
      label
      polygons {
         type
         points {
             x
             y
         }
       }
    }
}

Are there any estimates on query depth vs query speed?

In my view having a long graph to traverse is better. Also, a single string you have no fun.