China's unregulated AI models, created by companies such as DeepSeek and Alibaba, offer a feasible option to American proprietary systems

July 9, 2024, could be recalled as a day of disgrace for China'sartificial intelligencecommunity. On that day, the US-based company OpenAI, a leading global entity in AI model development,blocked developers in China- including Hong Kong and Macau - from utilizing its GPT models.

On the other hand, programmers from nations spanning from Afghanistan to Zimbabwe were granted entry, indicating OpenAI's implied view that its powerful models need to be protected from potential abuse by China, as well as Iran, Russia, and North Korea.

Currently, the situation has changed. With the launch of DeepSeek's free-to-use V3 large language model (LLM) in December 2024 and the release of DeepSeek's R1, an AI reasoning model that matches the performance of OpenAI's o1, the open-source initiative led by Chinese companies has created significant impact in Silicon Valley and Wall Street.

Are you curious about the most significant issues and developments happening globally? Find the solutions withSCMP Knowledge, our latest platform featuring curated content including explainers, FAQs, analyses, and infographics, presented by our acclaimed team.

The trend has not only triggered a surge of AI applications in China, but also reshaped the global AI scene, gaining the backing of developers around the world. Chinese open-source models offer a practical alternative to the closed systems promoted by American tech companies such as OpenAI and Google.

AI models that are open-source, meaning their source code and model parameters can be accessed by anyone for usage, alteration, and sharing, promote a cooperative method in the advancement of artificial intelligence.

Previously, open-source computer systems such as Linux were unable to surpass proprietary alternatives like Microsoft's Windows, but analysts believe that this time, China's free-to-use AI models present a significant threat to American counterparts.

Nvidia founder and CEO Jensen Huanghas commended China's advancements in open-source AI and stated a willingness to work with Chinese firms, as the semiconductor company gets ready to restart deliveries of its high-performance H20 chips to one of its major markets aftera significant development in Sino-US trade negotiations late last month.

Huang referred to large language models created by Chinese companies—such as DeepSeek, Alibaba Group Holding, Tencent Holdings, MiniMax, and Baidu—as "world-class" and essential for progress in global artificial intelligence. Alibaba owns the Post.

China's open-source AI initiative has acted as a "driver of global advancement," offering "each nation and sector an opportunity to participate in the AI transformation," he stated at the China International Supply Chain Expo in Beijing this week.

In comparison to the quick speed at which Chinese companies are launching their open-source models, the founder and CEO of OpenAISam Altman announced last weekend a postponement of its open-source LLMwhich was scheduled to debut in the coming days, according to safety issues and the requirement for further evaluation.

According to Kevin Xu, the founder of tech investment firm Interconnected Capital, for Chinese start-ups such as DeepSeek, embracing an open-source model proved to be a successful method to keep pace, as it enabled them to benefit from the input of a larger group of developers.

Following the release of ChatGPT by OpenAI in November 2022, Chinese open-source AI developers have achieved significant progress in enhancing their models. "Many open-source AI models from China are currently at or near the cutting-edge capabilities," according to Xu, competing with proprietary systems developed by leading US companies.

"The recent series of open-weight AI model launches highlights the increasing sophistication of open-source adoption and participation in China," he added.

DeepSeek's latest R1-0528is ranked as the top-rated open-source model in a study conducted by AI consulting firm Artificial Analysis, coming second only to models developed by Elon Musk's xAI, OpenAI, and Google. Several other Chinese companies, including Alibaba, MiniMax, and Moonshot AI, also showed strong results in these evaluations.

The sophisticated capabilities of Chinese models have not escaped the attention of users.

By mid-July, DeepSeek had a 24 percent stake in OpenRouter, an international platform for AI models, positioning it as the second-largest model creator, narrowly following Google, which held a 37 percent share.

In the meantime, Alibaba's Qwen model series has emerged as the world's largest open-source AI ecosystem, with more than 100,000 derivative models developed on its foundation, exceeding the Llama community from Meta Platforms, as reported by Hugging Face, the leading global open-source AI community.

China's broad open-source environment includes models with parameters ranging from 1 billion to 1 trillion, utilized across industries such as smart manufacturing and digital administration, as stated by Zheng Xiaolong, a researcher at the State Key Laboratory of Multimodal Artificial Intelligence Systems, linked to the Chinese Academy of Sciences.

The blending of technological advancement and industrial requirements has led to a distinct development approach in China, where practical needs spurred innovation and open-source communities accelerated industry expansion, Zheng noted in People's Tribune, a government-supported magazine.

The open-source initiative in China mirrored a growing push for "technological equity," according to him, opposing the prevalence of closed-source systems.

One factor behind Alibaba's decision to make its Qwen models open-source was that it "makes AI accessible to more people" and "encourages widespread use," which would benefit the company's cloud computing operations,Chairman Joe Tsai mentioned last month.

The open-source approach enabled Chinese companies to draw in both local and global developers, expanding their influence and usage, according to Ray Wang, research director for semiconductors, supply chain, and emerging technology at the consulting firm Futurum Group.

For instance, Chinese home appliance companies such as Midea Group and Haier are incorporating DeepSeek technology into their televisions and refrigerators, whereas American firms like Nvidia and Amazon.com provide users with access to DeepSeek's models.

Liu Zhi, the founder and CEO of the Chinese headphone company Oleap, mentioned that the open-source aspect of DeepSeek models "enables us to fine-tune them directly or perform further training and model compression".

Liu mentioned that due to DeepSeek's affordable pricing, the expense of employing AI to create meeting summaries through Oleap's smart headphones decreased by more than 80% following the integration of an API powered by DeepSeek's R1 model in February.

The achievements of DeepSeek led US companies such as OpenAI to reconsider their approaches, according to Jimmy Hu, who heads AI operations at Phoenix, a decentralized AI infrastructure firm based in Shanghai.

To remain ahead of the competition, US companies were anticipated to continue launching paid models that were "competitive or slightly better" than offerings from open-source alternatives such as DeepSeek, he mentioned.

In China, firms that once followed a closed-source model, like Baidu and MiniMax, are also reevaluating their strategies, as noted by Adina Yakefu, an AI researcher at Hugging Face.

"China's AI companies are making a collective move towards open source, which goes beyond mere symbolism; it indicates an increasing agreement that open source speeds up development, fosters trust, and enhances global reach," Yakefu stated.

China's extensive pool of skilled individuals and growing governmental backing have also contributed to the country's rising open-source movement, according to Bao Yungang, deputy director of the Institute of Computing Technology at the Chinese Academy of Sciences.

As of last year, official data cited by state broadcaster China Central Television indicated that China was home to 9.4 million software developers.

The nation provided 17 of the top 100 open-source software programs, positioning it as the second-largest contributor globally, as stated in a collaborative report by the China Communications Standards Association and the Cloud Computing Standards and Open-Source Committee.

China has also incorporated open source into its national strategy.

The 14th Five-Year Plan emphasized the significance of open-source development, detailing initiatives to establish local open-source AI communities, promote open-source innovations globally, and develop AI systems for gathering public data.

This initiative, driven by the state, has motivated local businesses and academic institutions to participate in the campaign, as stated by Bao.

With the growing global recognition of Chinese open-source AI models, Beijing's AI goals and the nation's AI offerings have come under political examination, especially from Western nations.

For instance, DeepSeek's chatbot has faced bans or restrictions in nations such as South Korea, Australia, Germany, Italy, and the Czech Republic due to issues related to data security.

Xu from Interconnected Capital stated that these actions were driven by political reasons rather than technical value. "Applying a national label to open-source code or models doesn't make technical sense," he mentioned.

Some have pointed out the worldwide advantages of companies—no matter their country of origin—exchanging their research and working together to promote AI development.

Don't overlook the wide-ranging global effects of open source," said Huang from Nvidia during his visit to Beijing. "Not only have open-source models supported the Chinese environment, but they are also benefiting systems worldwide.

More Articles from SCMP

Is Gibran from Indonesia being 'exiled' to Papua? New post raises speculation

Nvidia's chief executive holds talks with China's commerce minister regarding collaboration in artificial intelligence and international investment.

EU imposes penalties on two Chinese banks for supporting Russia

Why the family feud at Wahaha is a serious issue for Chinese businesses

This piece was first published in the South China Morning Post (www.scmp.com), a top news outlet covering China and Asia.

Copyright (c) 2025. South China Morning Post Publishers Ltd. All rights reserved.

Post a Comment