This repository compiles results of the unicorn test for various open-weights models.
This is not a good test. It is not even meant to be a good test.
However: Nobody is gaming it deliberately, because it is too silly. So it might be a better test than alternatives.