One year after DeepSeek, Chinese AI firms from Alibaba to Moonshot race to release new models

BEIJING — The race among Chinese technology firms to unveil new artificial intelligence models is gaining momentum as they seek to rival U.S. giants like OpenAI, Anthropic, and Google. This surge follows the disruptive launch of DeepSeek, a Chinese AI chatbot that significantly undercut OpenAI's ChatGPT in terms of both usage fees and production costs, stirring debates about the effectiveness of U.S. tech restrictions on China. On Tuesday, Moonshot AI, a startup based in Beijing, introduced its Kimi K2.5 model, featuring advanced video-generation and agentic capabilities that reportedly surpass the leading AI models from the U.S. Agentic AI refers to systems that can perform tasks independently on behalf of users, with the ultimate aim of achieving greater autonomy with minimal user input. This announcement follows closely on the heels of Moonshot's previous release of the K2 model just three months prior. Earlier in the day, Alibaba, the e-commerce behemoth, unveiled its generative AI model, Qwen3-Max-Thinking, which reportedly excels in creating text, images, or videos based on user instructions. Alibaba asserts that its new model outperformed significant U.S. competitors in a benchmark test dubbed "Humanity's Last Exam," and is capable of efficiently selecting the optimal AI tool for various tasks while utilizing past interactions for context, all at minimal additional costs. In a recent move, Z.ai launched a free version of its GLM 4.7 model, although it was forced to limit new sign-ups for its AI coding tool shortly after due to overwhelming demand that strained its computing resources. Additionally, shares of Baidu, another major Chinese tech player, surged to their highest in nearly three years following the debut of Ernie 5.0, its latest generative AI model, which Baidu claims outperforms Google's Gemini-2.5-Pro. Demis Hassabis, CEO of Google DeepMind, commented this month that Chinese AI models might be trailing U.S. developments by only a few months. Chinese firms are also making strides in promoting their homegrown technologies, often at lower costs, which could appeal to emerging markets. Unlike many leading U.S. AI models, Chinese alternatives tend to be open-sourced, allowing for free or low-cost access and customization of the underlying code. Alex Lu, founder of LSY Consulting, noted, "The hope is that countries outside of China will adopt these models, leading to a proliferation of applications built on them." Microsoft recently highlighted that DeepSeek's usage in Africa is estimated to be two to four times higher than in other regions, indicating a growing interest in Chinese AI solutions. The real value of AI, however, may lie in its integration with existing platforms. Ivan Su, a senior equity analyst at Morningstar, pointed out that focusing too much on AI benchmarks can detract from its potential when embedded in popular gaming or entertainment ecosystems, such as those operated by Tencent. Tencent announced plans to distribute 1 billion yuan (approximately $140 million) in cash rewards through its Yuanbao AI chatbot app during the Lunar New Year festival, reminiscent of its successful "red envelope" campaigns that helped establish WeChat as a leading mobile payment app. Other companies, including ByteDance and Baidu, are also launching AI-related promotions to engage users with their applications leading up to the holiday. Alibaba has updated its Qwen AI app to facilitate shopping, food ordering, and payments without leaving the interface, leveraging its e-commerce platforms like Taobao. With over 100 million monthly active users, this integration could enhance revenue for Alibaba, assuming that increased use of Qwen translates to higher shopping activity on Taobao, as suggested by Lu. Ultimately, the competition among Chinese tech firms appears to be driven more by user engagement than by the pursuit of the most advanced AI models.