Why Is China Choosing Open-Source AI? Diverging Strategies as U.S. Closes and China Opens Up

Both Kimi K3 and Qwen 3.8 Released as Open Source

Possible Alternative to Closed U.S. AI Models

Chinese artificial intelligence (AI) companies are pursuing a divergent strategy from the United States, which is focusing on protecting and treating cutting-edge AI models as strategic assets, by opening up their advanced AI models. As Chinese companies make even high-performance models available for business work and software development, competition between the U.S. and China over the global AI ecosystem and technological standards is expected to intensify.

Yonhap News

According to industry sources on July 20, Chinese big tech company Alibaba plans to release its AI model ‘Qwen 3.8 Max’, the latest in the Qwen lineup, using an open weights approach. This model, which has 2.4 trillion parameters, will make the weights shaped in the AI training process public. This allows companies to run the model on their own servers or modify it for their specific purposes.

Recently, Kimi K3, an open-source model developed by Chinese AI startup Moonshot AI, was released. It is designed for handling massive software repositories and long-duration coding tasks. On July 17 (local time), Chinese President Xi Jinping emphasized in a keynote speech at the World Artificial Intelligence Conference (WAIC) that “the advancement of AI should not be a solo performance by one country, but rather a symphony achieved through international cooperation.”

This strategy is interpreted as an attempt to spread derivative models and services based on open-source Chinese models, taking the lead in global technology standards and securing dominance in the AI ecosystem. With open-source models, users can directly download them and utilize, modify, and validate the models to fit their own data and services. Given their performance is on par with state-of-the-art American models but comes at a lower price, such models can be an option for small and medium-sized enterprises and emerging countries facing significant cost pressures.

In particular, these open models could serve as alternatives to those of the United States, which has recently shifted to operating closed AI models. While closed models allow developers to centrally manage service quality and safety mechanisms, they also carry the risk of being subject to platform operators’ policy changes, such as price increases or service suspensions. Although the U.S. once led the open-source ecosystem through Meta’s Llama and Google’s Gemma, it now recognizes advanced AI technology as a national strategic asset and is implementing a closed strategy, including tightening export controls.

According to the Financial Times (FT), the open-source-led Kimi K3 is seen as a threat to U.S. AI companies like Anthropic and OpenAI, which are keeping their most advanced models proprietary. Anastassios Angelopoulos, co-founder of AI evaluation platform Arena, commented, “Chinese open-source models could be a turning point that overtakes America’s closed models.”

However, a limitation of open-source models is that, since they can be freely downloaded and modified, it is difficult to control their stability, and accountability is unclear when problems arise. Jin Ho Park, Professor of Computer Science and AI at Dongguk University, explained, “The speed and performance of AI models have reached saturation, and there will be fierce competition between the U.S. and China.” He added, “A key limitation of open source is that it is unclear who is responsible when accidents or bugs occur.”

This content was produced with the assistance of AI translation services.