The Information suggests that Alibaba’s experience demonstrates Chinese companies’ potential to generate global impact by focusing on open-source AI models. Alibaba Group is taking a leading position in China’s open-source AI field, with its Qwen model surpassing Meta’s Llama in multiple benchmarks.
From internal skepticism to comprehensive adoption, Alibaba’s market influence continues to expand. According to The Information, internal business units were initially unconvinced by the Qwen model, with some teams continuing to use other companies’ AI models like Meta’s Llama as of 2024. Some Alibaba applications also opted for DeepSeek’s R1 model to support their AI functions. However, the situation changed significantly thereafter. Alibaba not only achieved full internal transition to Qwen but also gained widespread external recognition, becoming China’s largest open-source AI model provider. Some enterprise users even prefer Alibaba due to its broader model portfolio, including smaller models with lower operating costs. As of January this year, over 290,000 customers across various industries including automotive, healthcare, education, and agriculture have adopted Qwen models. Some AI application startups now choose Alibaba’s models over Llama when building software. Meanwhile, Alibaba Cloud is working to enhance Qwen’s global influence. In Japan, Tokyo-based AI developer Abeja has used Qwen to develop multiple large language models specifically designed for Japanese. The rise of China’s open-source AI sees Alibaba leading the development of large models. The Information analysis indicates that the success of Qwen and DeepSeek demonstrates Chinese companies beginning to lead the U.S. in open-source AI, an important front in the international AI competition. Martin Saerbeck, co-founder and CTO of Singapore’s Aiquris, stated in an interview: ‘Focusing on open-source AI models could enable Chinese companies to achieve global impact. Popular open-source models can leverage the collective knowledge of global developers and researchers.’ NVIDIA CEO Jensen Huang also remarked during a recent earnings call that DeepSeek and Alibaba’s Qwen are ‘among the best open-source AI models.’Notably, NVIDIA’s AI research team even used an Alibaba open-source model as the foundation when developing its Cosmos-Reason1 model for robotics and autonomous vehicles.
The Information conducted an in-depth analysis of how Alibaba has achieved a leading position in the open-source AI field. The report points out that this serves as a lesson for US tech giants like Amazon, Microsoft, and Google, as these companies operate in a more centralized manner than Alibaba. Alibaba’s leadership in open-source AI is partly attributed to its unique organizational structure. The company’s decision to allow different business units to operate autonomously, though initially intended to address corporate spin-off plans that ultimately didn’t materialize, unexpectedly prompted AI engineers to work harder to enhance model appeal. Engineers realized that if they couldn’t persuade internal business units to use the Qwen model, they wouldn’t be able to win external customer trust either. According to The Information, DeepSeek’s success put significant pressure on the Qwen team. Two informed sources revealed that even Jack Ma, Alibaba’s founder who stepped down from executive and director roles six years ago, frequently inquired about Qwen3 development progress from Alibaba Cloud CTO Jingren Zhou. In the final week before Qwen3’s release, team members took turns resting on mattresses under their desks. In late April this year, Alibaba finally released Qwen3, a suite comprising eight different models of varying scales and specifications, all open-source. Alibaba CEO Eddie Wu stated during the earnings call that the company is firmly committed to open-source AI. Multiple versions of Qwen3 surpassed Meta’s latest Llama 4 model on AI model leaderboards such as LiveBench and Artificial Analysis. Although DeepSeek released an updated R1 version last week that again outperformed Qwen3, Alibaba’s own AI products have shifted from DeepSeek to using Qwen. While Alibaba’s business units continue to operate independently, Qwen’s growing importance is helping bring them closer together. Many teams from various business units are now discussing plans with Alibaba Cloud to develop more powerful AI Agents powered by Qwen3. Alibaba’s experience demonstrates that Chinese companies have the potential to generate global impact by focusing on open-source AI models and may reshape the global AI software ecosystem.Meanwhile, China’s domestic open-source models will accelerate the proliferation and application of AI within the country.
As Alibaba CEO Eddie Wu Yongming stated, the company firmly believes that fully open-sourcing Qwen3 will drive innovation and new application development among developers, startups, and various enterprises. This article originates from the WeChat public account ‘Hard AI’. For more AI frontier insights, please visit the source. Risk warning and disclaimer: Markets involve risks, and investments require caution. This article does not constitute personal investment advice nor considers individual users’ specific investment objectives, financial situations, or needs. Users should assess whether any opinions, views, or conclusions herein align with their particular circumstances. Investments made accordingly are at one’s own responsibility.
