[kmacdough]

I too am skeptical of these models, but it's a reasoning focused model. As a result this isn't a very appropriate benchmark.

Small models are going to be particularly poor when used outside of their intended purpose. They have to omit something.