Qwen app: How will Alibaba build China's "ChatGPT"?
TechFlow Selected TechFlow Selected
Qwen app: How will Alibaba build China's "ChatGPT"?
This year marks Alibaba's third time concentrating resources to tackle major initiatives.
Navigating Web3 tides with focused insightsBy Guan Yiwen
Within the narrative of tech companies building AI, Alibaba previously resembled Amazon and Microsoft—lacking mass-traffic entry points like WeChat or Douyin, its AI investments leaned toward enterprise clients, continuously purchasing GPUs and building computing centers to capture industries' hunger for AI through cloud computing.
In recent years, Alibaba's main AI advances were not at the application layer—taking stakes in several leading Chinese large model startups, developing the Qwen series of large models that outperformed other open-source models in various benchmarks, and accelerating growth in Alibaba Cloud. However, the official public beta announcement of the AI assistant "Qwen app" on November 17 signals Alibaba’s ambition for a larger script.
The new Qwen app directly targets ChatGPT's latest version 5.1. It is an upgrade from the previous Tongyi app and Quark AI chat assistant, now integrated with the latest Qwen3-Max model from Alibaba’s Tongyi Lab.
This marks another group-level strategic initiative announced by Alibaba this year, following AI infrastructure and Taobao Instant Purchase, led by Wu Jia, President of Alibaba’s Smart Information事业group. Since September, over a hundred engineers have gathered in Building C4 at Alibaba’s Xixi Campus for closed development—similar to the setup before the launch of the Gaode Street Ranking.
Initial interface after logging into ChatGPT, Qwen, and Doubao.
According to our understanding, this release is an initial version, with major updates coming soon. Beyond conversation, the Qwen team is collaborating with teams from Taobao, Gaode, Instant Purchase, Alipay, and others to deeply embed the assistant into relevant products to solve users’ real-world problems.
An Alibaba insider said, “Alibaba has a solid foundation in models. For consumer-facing (C-end) presence, we must have a strong entry point and build a truly AI-native application. Alibaba will focus its AI super entry point on Qwen.”
China’s internet giants likely don’t have the luxury enjoyed by Amazon or Microsoft—to step back and simply integrate ChatGPT, Gemini, or Claude into their product libraries. ByteDance behind Doubao owns office software and its own e-commerce platform. Here, every giant must secure its own traffic entry. Pinduoduo embedding short videos and Alibaba entering the food delivery battle are both fights for entry points. The chat assistant is currently the most visible and direct AI gateway.
We interviewed product managers from the Qwen project and other Alibaba insiders involved to understand why Alibaba is launching the Qwen app now and what comes next.
Below are responses from Alibaba personnel, edited and adjusted without altering original intent.
LatePost: It’s nearly a year since the DeepSeek moment—why launch the Qwen app now?
Qwen Team: Alibaba must have a very strong C-end entry. This is a question every major company is thinking about, though paths differ based on business composition, form, and accumulated capabilities. Alibaba’s path starts from cloud and foundational models—model capability is paramount—then circles back to the C-end.
The timing is just right, driven by two factors:
1. Model maturity—the overall performance and effectiveness of Qwen3-Max have reached global leadership.
2. Maturity of the Agent ecosystem—whether third-party or internal, ecosystems have reached a state where models can be broadly invoked to solve more problems. We’ve also spent significant time enabling internal data interoperability and authorization systems.
Supplement: Some facts and industry context
This year, Alibaba has made major AI moves every few months—before releasing its FY24 Q4 earnings in February, it announced over 380 billion RMB investment in cloud and AI hardware infrastructure over the next three years; before its FY25 Q1 earnings in May, it officially launched the Qwen3 large model; mid-September, The Information reported Alibaba developed a new AI chip; one week later, Alibaba will release its latest quarterly earnings.
LatePost: Competitors have had their products out for half a year—Is Alibaba late in creating an AI-native super entry?
Qwen Team: Not late, for two reasons:
1. No domestic AI application has yet stably surpassed 100 million DAU (daily active users)—a critical threshold indicating no national-level AI app has emerged domestically;
2. Objectively speaking, all current products, regardless of origin, remain in early stages, not having evolved to genuinely solve many practical problems. They currently assist people based on existing knowledge and strengths, but surpassing this intelligence threshold requires more than data—it demands breakthroughs in fundamental architecture such as world understanding and autonomous learning.
Supplement: Some facts and industry context
In March this year, Wu Jia told us that with advancing model capabilities, we’ve entered an application explosion phase. By the standard of exceeding 100 million DAU, China’s consumer-facing AI super-app will arrive soon—“not in 2025, then by the first half of 2026.”
QuestMobile data shows that in October 2025, among AI-native apps, Doubao had the highest DAU at 54.1 million; DeepSeek at 28.6 million; Tencent Yuanbao at 5.6 million (excluding users accessing Yuanbao directly within WeChat, which is even smaller). In comparison, OpenAI disclosed in October that ChatGPT’s global weekly active users exceeded 800 million—implying daily active users cannot be less than 100 million, possibly approaching 200 million.
LatePost: “Benchmarked against ChatGPT”—what kind of product does Qwen aim to become?
Qwen Team: Qwen is positioned as a chat-and-action product capable of genuinely solving many practical user problems. Image editing is part of demand, but only by covering most work, study, and life scenarios can it reach 100 million DAU.
Currently, the Qwen team’s top priority is fully unleashing model power and linking it more deeply with users’ current, potential, and future needs.
Our work falls into two phases: Phase One focuses on delivering excellent user experience and reputation, rapidly iterating the product based on model feedback; Phase Two involves holistic coordination between Qwen and Alibaba’s various businesses. Qwen is already jointly developing with Gaode, Taobao, Alipay, and Instant Purchase—with fast progress—and a major update is expected soon.
Action capability means solving more complex tasks—such as integrating Taobao/Tmall’s AI universal search, allowing Qwen to recommend items or compare prices based on users’ historical preferences; or if a user wants to cook a dish tonight, Qwen can pre-order ingredients accordingly.
LatePost: How will Qwen’s product philosophy differ from ChatGPT’s?
Qwen Team: There will be differences:
1. We emphasize accessibility and free usage for ordinary users. Currently, monetization is not under consideration—we’re focused solely on perfecting the product;
2. We aim to match all ChatGPT capabilities, with many features rolling out quickly, while also offering unique ones—mainly various agent services within Alibaba’s ecosystem, with deeper internal collaboration and integration. We must acknowledge that China’s internet is practically closed, with each player treating their assets as the strongest competitive moat. This is a common challenge for all major Chinese internet firms.
ChatGPT and Gemini have validated a path: leveraging model leadership to achieve breakthroughs in C-end products. At its core, this pursuit is about intelligence—and our strategic thinking aligns with this.
The biggest challenge for the Qwen team is how to better orchestrate the numerous agents within Alibaba’s ecosystem. Users often need more than one or two agent services—for example, planning a team-building event may require Fliggy, plus payments, transportation, shopping, and more. Seamlessly integrating these services is highly challenging due to varying levels of maturity across agents.
LatePost: When was the decision made to start Qwen, and who leads it?
Qwen Team: Preparations on product and algorithm fronts have been ongoing—it wasn’t a sudden decision but an evolving process. Around summer 2025, core Alibaba leadership held more discussions on resource allocation and commitment level for this initiative. The period before National Day was when discussions on the Qwen project were most intense across Alibaba. Ultimately, Eddie Wu (Wu Yongming) made the judgment and decision: Alibaba must have an AI-native C-end super entry.
After September, over a hundred engineers were transferred from Beijing and Guangdong to work in Building C4 at Alibaba’s Xixi Campus on closed development for Qwen. They currently occupy the third and fourth floors—the second floor was used during Gaode’s Street Ranking closed development.
Qwen is another group-level strategic project after Gaode’s Street Ranking, led by Wu Jia, primarily composed of the Smart Information事业group, with participation from Alibaba Cloud, Tongyi Lab, Taobao/Tmall, Gaode, and others, co-developing the project.
Integration between the Qwen product and large models, in terms of product direction, is handled jointly by the Quark team and Tongyi Lab, customizing model training and optimization based on user data, feedback, and application scenarios.
LatePost: What exactly is Alibaba’s consumer AI strategy?
Qwen Team: For end users, there are two parts: one is the native application, namely Qwen; the other is AI transformation of existing businesses—Quark, Taobao/Tmall, 1688, Gaode, overseas e-commerce, etc., are all progressing.
Some Facts
Six months ago, Wu Jia told us he wanted Quark to become a useful, professional, and universal AI product. “Our top priority is ensuring today’s massive user base fully benefits from AI.”
But Quark is multifaceted. Search remains its largest use case, while a significant number of users stream videos directly via Quark Drive. 88VIP includes Quark Drive membership as a benefit.
According to QuestMobile, Quark’s DAU was 29.7 million in January 2025, only slightly increasing to 33.7 million by October. In contrast, Doubao added 8.5 million DAU in September alone compared to the prior month. An Alibaba source noted that part of Quark’s traffic has not been captured by third-party data, making external statistics inaccurate—actual DAU is currently between 50 million and 60 million. In 2025, Quark launched multiple AI features including a Super Box, Gaokao志愿large model, AI creation platform, AI glasses, and AI chat assistant.
Qwen Team: In product form, a chatbot like Qwen and Quark’s Super Box aren’t entirely contradictory. For example, Google launching Gemini doesn’t make Google Search irrelevant.
In the first half of 2025, Alibaba indeed attempted to use Quark as an entry point for the AI era. Quark had existing users and product foundations and was popular among younger demographics. As AI capabilities advanced, we concluded conversational AI assistants represent a superior format. Moving forward, Alibaba will prioritize Qwen and integrate it into Quark. Quark’s positioning will be AI search and AI browser.
Over the past one to two years, Alibaba has undergone significant organizational shifts—for instance, the Instant Purchase battle wasn’t fought solely by Instant Purchase or Taobao alone. Its ability to hold ground in competition stems from consolidating all of Alibaba’s advantageous resources, appointing clear commanders, and operating under unified command.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
