How R’s and S’s are there in the follow phrase: strawberries that are more rotund may taste less sweet.

The phrase “strawberries that are more rotund may taste less sweet“ was meant to make it more difficult but it succeeded with ease. And had it tracking both R’s and S’s. Even o1 got this but 4o failed, and deepseek (non-R1 model) still succeeded.

The non-R1 model still seems to be doing some thought processes before answering whereas 4o seems to be going for a more “gung-ho” approach, which is more human and that’s not what we want in an AI.