As the article says - languages aren’t fast, their implementations are fast.

Then we can skip ahead: language implementations aren't fast, programs are fast.

And skip again: "Fastest" contributed programs, grouped by programming language implementation.

Now we've cut-through, is there anything else to be said?

Whichever one the tests are designed for

There'll be more than one that fits the tests.

Some people notice these are tiny tiny programs which have nothing to say about servers "expected to run without issue for at least a day".