Where did you find the Nim example? It's by far the strangest fibonacci implementation I have seen (and I am a Nim developer).

Anyway, I translated the Go example to Nim for a better comparison:

    proc fib(a: int): int =
      if a < 2:
        return a
      return fib(a - 1) + fib(a - 2)

Many of these programs do not get tail-call elimination (e.g. I think most languages can not automatically optimize the equivalent of "return f(n-1) + f(n-2)", if they do tail-call elimination at all). The nicest ones do not bother with negative n, some others don't. IMHO not a very useful set of comparisons.

As I mentioned elsewhere, in python the correct way would be to add the @functools.lru_cache decorator. That way you get the caching behavior of the tricky to write methods, with the clarity of the obvious recursive solution.