qwq 32b seems good at reasoning. but both qwen and deepseek teams are doing bad in terms of truth. if it is getting smarter but also detaching from truth it is becoming concerning. (smart+truthful ok for me).