@d-healey said in WDF Model is not working in Faust:
Interesting that it gave you the right answer,
because the 4o is guessing. You can ask it to run a python script that counts rs in a string and you'll get the correct answer. Pure inference in this case is mostly guesswork, because it works with tokens, not letters.
07eec81d-c908-44ee-b8a3-c2a703e8f569-image.png
The o1 has reflection so it can catch itself doing this. Plus, this test comes up so often that I wouldn't be surprised it already made its way into the training data.