智谱公司发布的GLM 5.2模型恰逢华盛顿对Anthropic旗下顶级模型实施监管干预，时机巧妙地推动了大量用户转向GLM。知名科技投资人马克·安德森等业内人士纷纷表示，GLM 5.2是首个能「匹敌甚至超越」美国同类产品且「毫无妥协」的中国AI模型。Coinbase执行长也指出，通过默认使用GLM 5.2和Moonshot的Kimi等中国开放权重模型，在token用量指数级增长的同时支出保持平稳。
开放权重模型的吸引力不仅在于价格优势——成本仅为美国闭源模型的五分之一——更在于可控性与隐私性。传奇晶片设计师吉姆·凯勒表示，他的公司Tenstorrent已转向使用GLM和Kimi K2进行编码，因为开放模型可以自行微调、部署在自有伺服器上，拥有更大的控制权。美国AI巨头指控中国模型通过「蒸馏」窃取技术，但蒸馏本身是AI领域广泛使用的技术，且当今领先的AI公司自身产品也是建立在大量未经授权的创作内容之上，这使得这种指控缺乏说服力。
作者认为，华盛顿若以安全为由限制开放模型将是一个错误，因为这不会让技术消失，只会将其推向公众监督之外。AI研究人员内森·兰伯特警告，若现在禁止开放模型而仅允许闭源模型在少数公司手中变得更强大，未来将面临更大的问题。这场「中国冲击」已经直接威胁到支撑美国经济乐观预期的AI公司商业模式，美国应发挥算力优势，建设自身强大的开放权重生态系统，而非采取闭门恐慌式的应对。
Zhipu's release of its most advanced model, GLM 5.2, coincided with Washington's ad hoc regulatory moves against Anthropic's top Fable 5 and Mythos 5 models, triggering a wave of migration toward the Chinese alternative. Prominent figures including tech investor Marc Andreessen and the Coinbase CEO have publicly praised GLM 5.2 as the first Chinese AI model to rival American offerings without compromise, while noting that defaulting to open-weight Chinese models enables exponential growth in token usage at flat cost.
The appeal of open-weight models extends well beyond pricing, which can be five times cheaper than proprietary American systems. Legendary chip designer Jim Keller explained that his company switched to GLM and Kimi K2 for coding because these models can be fine-tuned, self-hosted, and fully controlled on private premises, offering genuine privacy. US incumbents' accusations of illicit distillation ring hollow given that distillation is a standard AI technique and that today's leading AI companies built their own products on millions of creative works often used without permission or payment.
The column argues that restricting open models on safety grounds would be counterproductive, pushing the technology outside public scrutiny rather than eliminating it. AI researcher Nathan Lambert warns that banning open models now while closed models grow dramatically more powerful in the hands of one or two companies would create a far greater problem. The author concludes that this Chinese AI shock is no longer hypothetical—it directly threatens the business models underpinning US economic optimism, and America must respond by building its own robust open-weight ecosystem, investing in research, and giving enterprises more control over their AI tools rather than resorting to closed-door panic.