中国AI初创企业的模型产品正在美国市场获得显著增长。Moonshot的Kimi K3、Z.ai的GLM-5.2等中国AI模型因价格低廉、性能接近美国顶尖模型而吸引了美国开发者和企业采用[1]。Kimi K3发布后一周内在美国的下载量增长了387%[1]，根据Sensor Tower的数据，该周下载次数超过930,000次[1]。
美国企业正在转向采用成本更低的中国AI模型。Coinbase等公司已经开始切换至这些中国模型以降低运营成本[1]。据报道，中国模型的成本为"每百万输出token几分钱"，而美国模型的成本则达到"30-50美元"[1]。Mozilla首席技术官Raffi Krikorian评价Kimi K3"看起来更灵敏"，相比Anthropic的Claude Fable[1]。这些中国模型正在与Anthropic和OpenAI等美国企业的闭源模型形成直接竞争[1]。
在商业表现方面，Z.ai去年收入激增132%至7.24亿元，但净亏损也跳升60%至47亿元[1]。Goldman Sachs在7月发布的研究报告中指出，中国AI模型已处于"关键阶段"[1]。此外，Trump政府指控Moonshot使用"隐蔽"方法基于Anthropic的Fable构建Kimi K3[1]。
Chinese AI startups are capturing market share in the United States with competitively priced models that deliver performance comparable to leading American alternatives. Moonshot's Kimi K3 and Z.ai's GLM-5.2 have attracted adoption from US developers and enterprises seeking cost savings, challenging proprietary models from companies like Anthropic and OpenAI.[1]
Kimi K3 experienced rapid uptake following its launch, with downloads in the United States surging 387 percent within the first week and exceeding 930,000 downloads during that period.[1] Mozilla CTO Raffi Krikorian described K3 as appearing "more responsive" compared to Anthropic's Claude Fable.[1] The cost advantage proves decisive for many organizations; according to analysts, Chinese AI models operate at "a few cents per million output tokens," while American counterparts command "$30 to $50 per million tokens."[1] Goldman Sachs identified Chinese AI models as having reached a "critical stage" in a July research report.[1]
Adoption extends to established US businesses transitioning to Chinese alternatives to reduce expenses.[1] Z.ai, one of the leading Chinese AI providers, reported revenue growth of 132 percent year-over-year to 724 million yuan, though net losses widened by 60 percent to 4.7 billion yuan.[1] The Trump administration has alleged that Moonshot employed "covert" methods to develop K3 based on Anthropic's Fable model.[1]