响马 (@xicilion) “我自己实际使用来看。gpt-5.4 还需要监督完成的工作，glm-5.2 已经可以自主完成了。” — TopicDigg

after testing glm 5.2 for almost a full day... there’s no way anyone still believes open weight models are 6/8 months behind i would say it’s one release away from seriously challenging gpt-5.5 and opus 4.8.. the scary part for openai and anthropic isn’t that glm already wins everywhere.. no, not yet.. they’re still ahead, but the gap doesn’t feel untouchable anymore...glm doesn’t need to beat them by a mile... it just needs to get close enough, because once intelligence feels close, price becomes the whole factor... and on cost we all know it’s not even comparable at all.. i think in a few months, running gpt or opus might feel like a premium luxury you only use for second opinions, architecture decisions, security reviews, or the really high-stakes stuff.. and for everything else open models might simply be good enough..and good enough + cheap enough, is all what everyone would want anyway..