There is a very simple explanation for why weaker models appear to kick sand in Fable's face: Fable cannot be benchmarked because of its batshit out-of-control refusal policy.

If it actually tackled all of the problems it was assigned, it would presumably kick Opus into the weeds.