I know what you guys think. Not open weight... but really, there's no way for us to tell. Except, there are some interesting hints here and there (check the attached screenshot).

I remember there was a website which mapped the LLM outputs in more robust way instead of simply comparing two outputs. If you're the author of that particular tool, please consider checking this model out and compare with the known model outputs to see which model family it belongs to, because I think this similarity here is kinda interesting.