[theropost]

I've been running this and a quantized version of the QwQ model and comparing the responses - so far, QwQ is working better, though that could change as I use them more and compare the outputs