在Anthropic的两款主要系统突然受到限制后，智谱发布了GLM-5.2模型，该模型与另外五款中国模型一同位列全球前十。中国开源人工智能模型正在将与美国同行的性能差距缩小至六个月或更短。这一崛起受到成本效益的推动；例如，GLM-5.2的使用成本仅为Claude Opus 4.8的约八分之一（12.5%）。尽管面临出口管制，智谱在全球人工智能任务使用量中仍位列第三。

然而，西方对数据安全和知识产权的担忧依然存在。Anthropic指控阿里巴巴使用2.4万（24,000）个虚假账户抓取其数据进行模型蒸馏，这一行为是被服务条款所禁止的。智谱的财务披露显示，其在2025年上半年的计算服务支出超过其营收的七倍（700%），这主要是因为租赁境外数据中心的计算资源以规避硬件出口限制。

来自ArenaAI的Anastasios Angelopoulos和来自Baseten的Charles O'Neill等行业专家指出，开发者对中国模型的采用率正在上升，特别是在人工智能代理领域。虽然部分开发者担忧监管风险，但其他专家警告称，压制美国开源软件可能会让中国主导该领域。随着Fable和Mythos等模型暂时退出市场，全球机构正在积极探索这些更便宜的替代方案，以避免对单一供应商的依赖。

Following the sudden restriction of Anthropic's two primary systems, Zhipu released its GLM-5.2 model, which ranks in the top ten globally alongside five other Chinese models. Chinese open-source AI models are narrowing the performance gap with US counterparts to six months or less. This surge is driven by cost efficiency; for instance, GLM-5.2 costs approximately one-eighth (12.5%) of Claude Opus 4.8. Despite export controls, Zhipu ranked third in global AI task usage.

However, Western concerns persist regarding data security and intellectual property. Anthropic accused Alibaba of using 24,000 fraudulent accounts to scrape its data for distillation, a practice prohibited by service terms. Financial disclosures from Zhipu reveal that its computational service expenditures in the first half of 2025 exceeded its revenue by over seven times, primarily due to renting computing resources from offshore data centers to circumvent hardware export restrictions.

Industry experts like Anastasios Angelopoulos of ArenaAI and Charles O'Neill of Baseten highlight the growing adoption of Chinese models among developers, especially for AI agents. While some developers fear regulatory risks, others warn that suppressing American open-source software could allow China to dominate. With the temporary withdrawal of models like Fable and Mythos, global organizations are actively exploring these cheaper alternatives to avoid reliance on a single provider.