I tried a few models, all seemed to get it (although I had never heard of that paradox).

That’s a very poor answer IMO. The point of the “paradox” is that non-black non-ravens are literally zero evidence about the color of ravens.

Like, the answer acknowledges the paradox but tries to resolve it by saying “well, it’s just very weak evidence”—which is actually denying that the paradox exists. In that sense, the answer is incoherent in exactly the way that many student papers I get written by LLMs are incoherent.