在旧金山举行的《世界其他地区》活动上，行业专家报告称，美国和欧洲开发者（包括Reflection、Aleph Alpha和Mistral）新发布的开放权重人工智能模型在性能和成本效率方面仍落后于中国同类模型。
Chinese Open-Weight AI Models Outperform Western Competitors Amid Global Adoption Push At a Rest of World event in San Francisco, industry experts reported that newly released open-weight artificial intelligence models from U.S. and European developers, including Reflection, Aleph Alpha, and Mistral, continue to trail Chinese counterparts in performance and cost efficiency.
受美国芯片出口管制推动，DeepSeek、腾讯和小米等中国企业开发出高效架构，其任务执行成本比美国领先前沿系统低50%至90%。
Driven by U.S. chip export controls, Chinese firms such as DeepSeek, Tencent, and Xiaomi developed highly efficient architectures that perform tasks at 50% to 90% lower costs than leading U.S. frontier systems.
低收入国家和国际开发者越来越多地采用中国开放模型，以规避对美国技术的地缘政治依赖，尤其是在美国当局限制外国访问特定专有模型之后。
Lower-income countries and international developers are increasingly adopting Chinese open models to avoid geopolitical dependency on American technology, particularly after U.S. authorities restricted foreign access to select proprietary models.
尽管具有技术优势，中国AI实验室仍面临算力短缺和变现挑战，同时正寻求全球合作伙伴关系。
Despite their technical edge, Chinese AI labs face persistent compute shortages and monetization challenges while pursuing global partnerships.