This year, Alibaba is focusing its efforts on major initiatives for the third time.
Written by: Guan Yiwen
In the narrative of tech companies building AI, Alibaba has previously resembled Amazon and Microsoft in the U.S. — lacking a universal traffic entry point like WeChat or Douyin, its AI investments have leaned towards enterprise clients, continuously purchasing cards to build computing power centers, and capturing the demand for AI across various industries through its cloud computing business.
In the past few years, Alibaba's main progress in AI has not been at the application level — it has invested in several leading Chinese large model startups, its self-developed Qwen series large models have outperformed other open-source models in various tests, and Alibaba Cloud has accelerated its growth. However, the AI assistant "Qianwen app" officially announced its public beta on November 17, indicating that Alibaba aims for a larger narrative.
The new Qianwen application is clearly positioned against the latest version 5.1 of ChatGPT. It is an upgrade from the previous Tongyi app and Quark AI dialogue assistant, integrating Alibaba's Tongyi Laboratory's latest Qwen 3 - Max model.
This is another strategic project announced by Alibaba this year, following AI infrastructure and Taobao Flash Sale, with Wu Jia, President of Alibaba's Intelligent Information Business Group, in charge. Since September, over a hundred Alibaba engineers have gathered in the C4 building of Alibaba's Xixi campus for closed development — similar to the launch of the Gaode Street Ranking.
Initial interface of ChatGPT, Qianwen, and Doubao after launch.
According to our understanding, the version being launched is a basic one, with significant updates expected soon. In addition to chatting, the Qianwen project team is also collaborating with teams from Taobao, Gaode, Flash Sale, Alipay, and other products to develop deeper integrations, aiming to solve users' practical problems.
An Alibaba insider stated, "Alibaba has a solid model foundation, and there must be a strong entry point for the consumer side. We need to create an AI-native application, and Alibaba will focus on Qianwen as the key entry point for AI."
Chinese internet giants may not have the luxury of choices like Amazon and Microsoft — stepping back to let ChatGPT, Gemini, and Claude integrate into their product libraries. Doubao's parent company, ByteDance, has both office software and its own e-commerce platform. Here, every giant must ensure they have a traffic entry point. The built-in short videos in Pinduoduo and Alibaba's entry into the food delivery battle are both competitions for entry points. The chat assistant is currently the most direct visible AI entry.
We interviewed the product manager of the Qianwen project and other Alibaba participants to understand why Alibaba is launching the Qianwen app now and what the subsequent strategy is.
Below are the responses from Alibaba-related individuals, edited and adjusted without changing their original meaning.
Late Point: It has been almost a year since the DeepSeek moment; why launch the Qianwen app now?
Qianwen Team: Alibaba definitely needs a very strong consumer entry point. This is a question every large company is considering and discussing, but the paths may differ based on each company's business composition, business form, and accumulated capabilities. Alibaba's path will certainly start from the cloud and foundational models, as model capability is the most important, and then return to the consumer side.
The timing is just right, driven by two reasons for us to focus on developing this native AI product, Qianwen:
The maturity of the model; the overall performance and effectiveness of Qwen3 - Max have reached a global leading level.
The maturity of the entire agent ecosystem; whether it's the third-party ecosystem or the internal ecosystem of the group, it has reached a state where the model can be widely called upon to solve more problems. We have also spent a lot of time handling data interoperability for internal business and the overall authorization capability.
Supplement: Some facts and industry situations
This year, Alibaba has made significant moves regarding AI every few months — before the release of the 24Q4 natural season financial report in February, it announced plans to invest over 380 billion yuan in building cloud and AI hardware infrastructure over the next three years; before the release of the 25Q1 natural season financial report in May, it officially launched the Tongyi Qianwen Qwen3 large model; in mid-September, The Information reported that Alibaba was developing a new AI chip; a week later, Alibaba will release a new quarterly financial report.
Late Point: Competitor products have been launched for half a year; is it too late for Alibaba to create a native AI super entry?
Qianwen Team: It's not late, for two reasons:
No AI application in China has consistently broken the 100 million DAU (daily active users) mark, which is an important threshold, indicating that at least there hasn't been a national-level AI application in the country yet;
Objectively speaking, no product in China has truly evolved to the point of solving many practical problems. Everyone is assisting people based on their own knowledge and advantages, but crossing this intellectual threshold requires more than just data; it relies more on breakthroughs in understanding the world and autonomous learning at the foundational level.
Supplement: Some facts and industry situations
In March of this year, Wu Jia told us that with the development of model capabilities, we have now reached an explosive stage for applications. If we use over 100 million DAU as a standard, a Chinese AI super application aimed at consumer users will definitely arrive soon, "not in 2025, but in the first half of 2026."
QuestMobile data shows that by October 2025, among AI native applications, Doubao had the highest DAU at 54.1 million; DeepSeek had 28.6 million, and Tencent Yuanbao had 5.6 million (not including users directly using Yuanbao in WeChat, which is even fewer). In comparison, OpenAI disclosed in October that ChatGPT had over 800 million global weekly active users, which implies that daily active users cannot be less than 100 million and may be close to 200 million.
Late Point: "Benchmarking ChatGPT," what kind of product does Qianwen aim to become?
Qianwen Team: Qianwen is positioned as a product that can chat and handle tasks, genuinely helping users solve many practical problems. Image editing is also part of the demand, but only by covering most of the issues users face in work, study, and life can we reach 100 million DAU.
Currently, the most important task for the Qianwen team is to fully unleash the power of the model and connect it with users' current, potential, and future needs at a higher level.
Our main work is divided into two phases: the first phase focuses on enhancing user experience and reputation, rapidly iterating the product based on the model; the second phase involves overall collaboration between Qianwen and various Alibaba businesses. Currently, Qianwen is already engaged in joint development with Gaode, Taobao, Alipay, Flash Sale, etc., with very rapid progress, and a major version update is expected soon.
The ability to handle tasks refers to solving more complex tasks, such as combining Taotian's AI universal search, where Qianwen recommends products based on past user preferences, even comparing prices; for example, if a user wants to cook a dish one evening, Qianwen can help the user purchase the corresponding ingredients in advance.
Late Point: How will the product approach of Qianwen differ from that of ChatGPT?
Qianwen Team: There will be differences:
We emphasize that ordinary people can use it anytime and for free. Currently, we are not considering charging and are focused on product quality;
We hope to have all the capabilities that ChatGPT has, many of which will be launched quickly, while also having features that ChatGPT does not, mainly various agent services within the Alibaba ecosystem. The level of internal business cooperation and collaboration will be deeper. We must acknowledge that the Chinese internet is actually closed, with everyone treating their assets as the strongest competitive barrier. This is a common issue for every major internet company.
ChatGPT and Gemini have validated a path where relying on leading models can achieve breakthroughs in consumer-side products. The essence of this is the pursuit of model intelligence, and our thought process is the same.
The biggest challenge for the Qianwen team is how to better connect the many agents within the Alibaba system based on scenarios. Many user needs do not rely on just one or two agent services; for example, if I want to plan a team-building event, the services needed may not only include Feizhu but also a range of capabilities like payment, travel, and shopping. How to organically and smoothly connect them is quite challenging, as the maturity of each agent's capabilities varies.
Late Point: When was it decided to start developing Qianwen, and who led the initiative?
Qianwen Team: Preparatory work on the product and algorithm level has been ongoing; it wasn't a one-time decision but an evolutionary process. Around the summer of 2025, discussions among Alibaba's core management became more frequent, determining how many resources and how much effort Alibaba would invest in this initiative. Before the National Day holiday, discussions about the Qianwen project were the most intensive across Alibaba. Ultimately, it was Wu Yongming's judgment and decision — Alibaba must have a native AI super entry point for the consumer side.
After September, over a hundred engineers were brought in from Beijing and Guangdong, and everyone worked on the Qianwen project in a closed environment in the C4 building of Alibaba's Xixi campus. The current team is located on the third and fourth floors, while the Gaode Street Ranking team was on the second floor during its closed development.
Qianwen is another strategic project for the group following the Gaode Street Ranking, with Wu Jia as the person in charge. The team primarily consists of the Intelligent Information Business Group, along with many other businesses participating, including Alibaba Cloud, Tongyi Laboratory, Taotian, Gaode, etc., all working together on joint development.
The integration of the Qianwen product and large model involves the Quark team and Tongyi Laboratory customizing model training and optimization based on the user data and feedback received, as well as various application scenarios.
Late Point: What is Alibaba's specific strategy for consumer-side AI?
Qianwen Team: For users, one aspect is the native application, which is Qianwen; the other aspect is the AI transformation of existing businesses, with Quark, Taotian, 1688, Gaode, and overseas e-commerce all working on this.
Some Facts
Six months ago, Wu Jia told us that he hoped Quark would become a useful, professional, and versatile AI product for users. "Our top priority is definitely to ensure that the large user base of Quark can enjoy the value brought by AI."
However, the Quark application encompasses a wide range of functions. The greatest demand is for search, and a considerable number of users also watch shows directly on Quark's cloud storage. One of the benefits of 88VIP is a complimentary membership for Quark's cloud storage.
According to QuestMobile data, Quark's DAU was 29.7 million in January 2025, and by October, this number had only slightly increased to 33.7 million. In contrast, Doubao saw an increase of 8.5 million DAU in September compared to the previous month. An Alibaba insider mentioned that part of Quark's traffic has not been marked by third-party data, leading to inaccurate external statistics; the actual DAU is between 50 million and 60 million. In 2025, Quark has been very active in AI, launching features like Super Box, college entrance examination volunteer large model, AI creation platform, AI glasses, and AI dialogue assistant.
Qianwen Team: In terms of product form, a chatbot like Qianwen and Quark's Super Box are not entirely contradictory. For example, it’s not that Google created Gemini, so Google Search is no longer important.
In the first half of 2025, Alibaba indeed hoped to use Quark to explore entry points in the AI era. Quark has its original user base and product foundation, and it is a product that young people are using. With the enhancement of AI capabilities, we believe that a conversational AI assistant is a better form. Moving forward, Alibaba will focus on developing Qianwen and integrate it into Quark. Quark's positioning is AI search and AI browser.
In the past year or two, Alibaba has undergone significant organizational changes; for instance, the Flash Sale initiative was not solely driven by Flash Sale or Taobao. Its ability to stand firm in competition is fundamentally due to the concentration of all of Alibaba's advantageous resources, with a clear commander and unified command.
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
