Not every framework presented in the benchmark perform the same amount of necessary protocol verifications. For instance, here's the take from Servant perspective - https://www.reddit.com/r/haskell/comments/6b3dlt/techempower...