申宇婧等：前沿大模型差距仅2.7%！中美AI竞争已并跑

编者按：围绕AI的竞合正在深刻重塑21世纪的大国力量对比。8月18日，中国人民大学重阳金融研究院宏观研究部副主任、副研究员申宇婧与实习研究员董冀萌在发表英文文章表示，中美两国应更多地在AI领域开展良性竞争、深化务实合作，促进AI技术造福于世界各国人民。这不仅契合两国人民的根本利益，也是国际社会的共同期待。现将中文译文及英文原文发布如下： （中文译文约1800字，预计阅读时间5分钟）

▲英文原文发表于《中国日报》2026年8月18日第8版

当今世界正经历百年未有之大变局，以大模型为代表的新一轮人工智能 (AI) 技术革命正加速演进，AI已从单纯的技术问题上升为重塑全球竞争格局、定义国家领导力的核心战略变量。作为引领新一轮科技革命和产业变革的战略性技术，AI正全方位重构生产方式、产业结构与治理模式。围绕AI的竞合正在深刻重塑21世纪的大国力量对比。

在数智时代，AI不只是一个技术问题，它从经济、治理、国家安全等不同维度，成为衡量国家综合实力和治理能力的关键指标，是关系到一国未来竞争力的核心要素。可以说，能否抓住AI发展的战略机遇，掌握AI发展的主动权，直接决定了一国在未来全球竞争中的地位，是国家竞争力提升不可或缺的核心引擎。

以美国为代表的主要国家近两年加紧制定促进AI发展的国家顶层战略规划，将AI深度嵌入到国家发展战略之中，举全国之力试图维系自身在AI领域的竞争优势。2025年7月，特朗普政府颁布《赢得AI竞赛：美国AI行动计划》，将AI定位为决定21世纪全球权力格局的终极战略制高点，强调在国际AI外交与安全领域加大芯片出口管制力度，维护美国的绝对领先优势。

2025年11月，美国启动“创世纪计划”，强调这是一项具有历史意义的全国性行动，将其紧迫性和雄心类比二战期间的曼哈顿计划，通过整合17个国家实验室的超算资源、科研数据与私营部门大模型能力，搭建统一的国家级科学算力平台。2026年7月，白宫发布《科学：新的黄金时代》报告，提出改革美国的科研体系，针对AI等具有重大战略意义的领域，应重启曼哈顿计划和阿波罗计划的模式，由政府确立国家目标，集中全国资源。

目前，从整体实力和水平来看，中美领衔全球AI发展的格局愈加明显。中美无论在算力规模、产业应用还是核心论文产出层面，都占据了全球绝大多数的资源与成果，其他国家难以在短期内撼动两者的领先地位。

在算力方面，中美两国在顶级高端芯片储备与算力中心规模上保持领先。牛津大学近期发布的《人工智能算力主权：跨地域、云服务商与算力加速器的基础设施管控》报告显示，全球算力工厂主要集中在中美两国，其中美国有26个，中国有22个。

在大模型方面，根据美国斯坦福大学2026年《人工智能指数报告》，美国在开发顶级AI模型方面仍处于优势地位，但中国正在迅速赶上。虽然美国在数量上保持领先，但中国的模型在质量上迅速缩小了差距。中美前沿大模型综合性能差距仅为2.7%，两国模型多次交替领先，已进入技术并跑阶段。

在人才、数据等支撑AI发展的资源要素方面，美国聚集了全球最多的顶尖研究人员，在高端原创人才吸引力上保持优势；中国则拥有全球规模最大的AI工程师储备，每年理工科专业毕业生超500万人。2025年中国数据生产总量达52.26ZB，占全球27.44%，数据资源规模位居全球前列。人才与数据的规模优势为中国AI长期发展提供了充足支撑。

AI技术是人类共同的事业，其良性发展关乎全人类的共同福祉。作为引领全球AI发展的两大力量，中美两国应超越博弈竞争框架，携手加强合作，共同应对AI领域的全球性挑战，促进AI向善发展。

首先，中美可共同支持“全球南方”国家的AI发展，构建开放包容的全球AI治理体系。当前，全球AI发展不平衡的问题依然突出，广大发展中国家在算力、数据、人才、技术等方面面临诸多瓶颈。中美两国应坚持共商共建共享原则，通过技术转让、能力建设、人才培养等多种方式，帮助发展中国家提升AI发展能力。推动形成普遍参与、普遍受益的全球AI治理规则，共同营造开放、公正、包容的AI发展环境。

其次，中美应共同防范AI带来的风险挑战，守护人类共同安全。AI技术的快速发展在带来巨大机遇的同时，也伴随着算法偏见、隐私泄露、伦理冲突等一系列风险挑战，甚至对国际安全和全球稳定造成一定的影响。面对这些全球性风险，中美两国应在AI安全治理、伦理规范、标准制定等方面加强对话合作，共同建立健全AI风险预警和应对机制，确保AI始终在安全可靠的轨道上行稳致远。

最后，中美要共同推动AI技术造福全人类，应对全球性挑战。AI为解决气候变化、公共卫生、粮食安全、能源短缺等全球性问题提供了新的工具和可能。中美两国可以在AI赋能绿色低碳转型、药物研发、优质教育资源普惠等领域开展务实合作，让AI技术成为解决人类共同挑战的有力武器。

历史经验表明，要让技术革命红利惠及全人类，离不开大国之间的良性互动与责任引领。中美两国应更多地在AI领域开展良性竞争、深化务实合作，促进AI技术造福于世界各国人民。这不仅契合两国人民的根本利益，也是国际社会的共同期待。

英文文章内容

A new round of the artificial intelligence revolution driven by large language models is accelerating at a rapid pace. AI has evolved far beyond a purely technical issue to become a core strategic variable reshaping the global competitive landscape. As a transformative technology spearheading the new scientific and industrial revolution, AI is comprehensively reshaping modes of production and industrial structures. Competition and collaboration centered on AI will help shape the balance of power among major countries in the 21st century.

Over the past two years, major economies have stepped up efforts to formulate top-level national AI strategies, embedding AI in national development blueprints and mobilizing nationwide resources to sustain their competitive edges in this field. In July 2025, the United States released Winning the AI Race: America's AI Action Plan, charting a decisive course to cement US dominance in AI. In November 2025, the US launched the Genesis Mission, described as a historic nationwide initiative comparable in urgency and ambition to the Manhattan Project during World War II. It integrates supercomputing resources, research datasets from several national laboratories and private-sector large-scale model capabilities to build a unified computing platform. In July 2026, the White House published the report Science: A New Golden Age, advocating an overhaul of the US research system.

China and the US are leading the global development of AI. According to core AI indicators in Stanford University's 2026 AI Index Report, China and the US account for the vast majority of global resources and outputs in computing capacity, industrial application and high-impact academic papers — a lead unlikely to be challenged by other nations in the near term. In computing supply, both countries maintain advantages in high-end chip reserves and the scale of computing hubs. An Oxford University study titled AI Compute Sovereignty: Infrastructure Control Across Territories, Cloud Providers and Accelerators, found that major AI computing clusters are concentrated in the two nations: 26 in the US and 22 in China.

Stanford's report also indicates that the US retains an edge in developing top-tier frontier models, yet China is closing the gap rapidly. While the US leads in quantity, Chinese models are catching up fast in quality. The comprehensive performance gap between leading Chinese and US frontier models stands at just 2.7 percent, with models from both sides alternating at the top of rankings since early 2025, marking a phase of parallel technological progress. Regarding supporting factors such as AI talent and data, the US still hosts the world's largest pool of elite AI researchers and retains advantages in attracting top original innovators. According to China's National Development and Reform Commission, China possesses one of the biggest cohorts of AI engineers, producing more than 5 million STEM graduates every year. In 2025, according to the National Data Resource Survey Report (2025) released by China's National Data Administration, China generated 52.26 zettabytes of data, accounting for 27.44 percent of the global total and ranking among the world's largest data resource holders. These scale advantages in talent and data lay a solid foundation for sustained long-term AI growth.

AI is a shared endeavor of humanity, and its sound development bears on the common well-being of all. As the two major forces driving global AI progress, China and the US can transcend zero-sum competition, strengthen their collaboration, jointly address global challenges arising from AI and advance the ethical and beneficial development of AI.

First, China and the US can work together to support AI cooperation between the Global North and the Global South as well as South-South cooperation to build an open and inclusive global AI governance system. Severe imbalances persist in global AI development, with many developing countries facing bottlenecks in computing power, data, talent and technology. Both sides should abandon zero-sum thinking, uphold the principle of extensive consultation, joint contribution and shared benefits, and help developing countries build their AI capacity through technology transfer, capacity building and talent training.

Second, China and the US can jointly guard against AI risks. While rapid AI innovation unlocks enormous opportunities, it also brings risks including algorithmic bias, privacy breaches and ethical conflicts, with potential implications for international security and global stability. Faced with these transboundary risks, China and the US can deepen dialogue and cooperation on AI safety governance, ethical norms and standard-setting, and jointly improve early-warning and response mechanisms to ensure AI evolves along a secure and reliable track.

Finally, the two countries can jointly harness AI to benefit humanity and tackle global challenges. AI offers new tools to address climate change, public health emergencies, food insecurity and energy shortages. The two nations can pursue pragmatic cooperation in AI-enabled low-carbon transition, pharmaceutical research and equitable access to quality education, turning AI into a powerful instrument for resolving shared challenges facing humankind.

History shows that the dividends of technological revolutions can reach all humanity only through positive interaction and responsible leadership among major countries. While pursuing healthy competition, China and the US should deepen practical cooperation to enable AI technologies to serve the well-being of people across the world. This serves the fundamental interests of both countries and meets the shared expectation of the international community.

欢迎

加入人大重阳社群

为增进与粉丝们的互动，让大家有更好的交流体验，人大重阳建立了“企业”“学术”两大社群，欢迎感兴趣的朋友扫描上方二维码添加人大重阳工作人员的微信（rdcy2013sph）备注姓名、单位、职务及想加入的群聊，审核通过后工作人员会将您拉入群内。

// 人大重阳

RDCY

中国人民大学重阳金融研究院（人大重阳）成立于2013年1月19日，是重阳投资向中国人民大学捐赠并设立教育基金运营的主要资助项目。

作为中国特色新型智库，人大重阳聘请了全球数十位前政要、银行家、知名学者为高级研究员，旨在关注现实、建言国家、服务人民。目前，人大重阳下设7个部门、运营管理4个中心(生态金融研究中心、全球治理研究中心、中美人文交流研究中心、中俄人文交流研究中心)。近年来，人大重阳在金融发展、全球治理、大国关系、宏观政策等研究领域在国内外均具有较高认可度。

别忘了“点赞”+“转发”