DeepSeek open-sources the DeepInfer inference engine.

中国人工智能公司深度求索公开了其推理框架DeepInfer的源代码。该软件已在GitHub上以Apache 2.0许可证发布，正是用于大规模部署语言模型DeepSeek-V4.1-Flash的核心系统，能够保证高性能与低延迟。此次开源恰逢特朗普政府宣布禁止美国联邦民事机构使用深度求索工具的次日。

中国商务部对此美方决定提出严正关切，指出此举违反了两国元首习近平主席与特朗普总统达成的共识，并要求撤销这一"歧视性措施"。北京方面已明确表示，将采取必要反制措施以维护中国科技企业的合法权益。

DeepSeek is a Chinese company that develops advanced artificial intelligence models, competing directly with giants like OpenAI. Its V4.1-Flash model is considered one of the most powerful and cost-effective on the market. The DeepInfer framework, now open source, is the software that enables the use of these "trained" models to respond to user requests (a phase called "inference"), efficiently handling large volumes of traffic. Its public release aims to reduce the computational costs and energy required for AI, making it more accessible.

Tensions between China and the United States on the technological front are not new, but they have intensified with the rise of AI as a strategic sector. The U.S. ban on using DeepSeek in civilian government agencies is motivated by cybersecurity concerns and the protection of sensitive data, given the potential ties of Chinese tech companies to the Beijing government. This clash is part of a broader technological and commercial war between the two superpowers, where control over emerging technologies like AI is considered crucial for national security and economic supremacy.