Off-topic question: I've read other comments or articles here recently, where people uses models such as GLM (from Openrouter for example) to perform rev. engineering tasks. On the other hand, I can see offerings for abliterated GLM. Besides, I can see comments in this thread about models like Anthropic's refusing to RE games, but it doesn't happen, apparently, with open weights models, even with official providers. So, what's the difference with this abliterated Chinese models?