Meta shifts strategy on open-source AI development, leaving the field to China's advancements in sophisticated models
In a significant shift, China has surpassed the United States in open-source artificial intelligence (AI) development, as of mid-2025. This achievement is the result of a coordinated, systemic approach that integrates infrastructure, talent cultivation, and open collaboration.
A Centralized Strategy for Success
The Chinese government's strategy is markedly different from the decentralized, fragmented approach in the United States. China has established a National Integrated Computing Network, pooling public and private computing resources, and promoting green energy-aligned data center expansion [1][4]. This infrastructure-first model has allowed China to make rapid strides in the field of AI.
Open-Source Models Fuel Innovation
China's AI industry leaders, such as Alibaba, ByteDance, Huawei, and Baidu, are actively releasing open-source AI models. This has fostered a broad developer and researcher ecosystem that accelerates innovation and talent development [1][2]. The open availability of source code and model weights facilitates rapid experimentation and application to industrial contexts, extending China's "technological soft power" on a global scale.
Global Cooperation and Influence
China promotes global AI cooperation with fewer conditions than the United States, which ties AI exports to political alignment. This open policy may enhance China’s influence in global AI governance frameworks while supporting broader international adoption of its open-source models [3].
The U.S. Approach
In contrast, the United States focuses more on proprietary large language models, enterprise AI, and semiconductor technologies. While it has increased federal AI funding and released a comprehensive AI action plan, its decentralized regulatory environment and energy policy shifts slow infrastructure growth relative to China’s tightly controlled and multiply coordinated investments [1][3][4].
Top Proprietary Models Remain in the U.S.
Despite China's advancements in open-source AI, the world's top proprietary AI models are still from frontier U.S. labs, according to a post published by Wu on DeepLearning.AI [5]. Meta Platforms, the parent company of Facebook, is known for its Llama family of open-source AI models.
Notable Developments in China
One of the most significant developments is Alibaba's Wan 2.2 video tool, which is the first open-source video generation model incorporating the Mixture-of-Experts (MoE) architecture [2]. This tool is designed to help users unleash film-level creativity.
Another notable development is Zhipu's GLM-4.5, a reasoning and video model that secured third place globally and first place among both domestic and open-source models based on the average score across "12 representative benchmarks" [2]. Zhipu's GLM-4.5 is described as China's most advanced open-source MoE model, indicating a growing competitive edge for China in the development of open-source AI models.
The Future of AI
Andrew Ng, a renowned computer scientist, has praised China's open AI ecosystem for fostering competition among companies in a "Darwinian life-or-death struggle" to advance foundational models [6]. The successive launches of open-source models by Chinese companies suggest continued advancement in China's open-source AI ecosystem [7].
In conclusion, China's surpassing of the United States in open-source AI is a testament to a government-led integration of infrastructure development, open-source model release, concentrated talent cultivation, and international collaboration policies, resulting in a formidable AI ecosystem and expanding global influence [1][2][3][4].
References: 1. Towards Data Sovereignty: China's National Integrated Computing Network 2. China's AI Advancements: Alibaba's Wan-2.2 and Zhipu's GLM-4.5 3. China's AI Strategy: Open Cooperation vs. U.S. Export Restrictions 4. China's AI Infrastructure: A Centralized Approach 5. The State of AI: Open vs. Proprietary Models 6. Andrew Ng Praises China's Open AI Ecosystem 7. China's Open-Source AI Models: A Sign of Continued Advancement
- In line with the centralized strategy, China's AI industry leaders are not only focusing on the development of open-source AI models but also integrating these models with advanced technologies such as science, technology, and artificial intelligence.
- This competition among Chinese companies, driven by the open AI ecosystem, is pushing the boundaries of innovation in fields like education and art, creating opportunities for the integration of AI with various disciplines, fostering a comprehensive and versatile AI development.
