Qwen's story of using open source to defeat GPT is something Qianwen App wants to do again.
Alibaba is preparing to use the same strategy that led to Qwen's model 's victory over GPT again.
In 2018, OpenAI released its first model, GPT1, gaining a technological advantage, and subsequently became increasingly closed off. On the other side of the ocean, Alibaba began research on large models around the same time, and when it launched "Tongyi 1000 Questions" in 2023, it chose a completely different path: directly open-sourcing the model, allowing developers to use, improve, and integrate it for free.
This strategy has allowed Qwen to gradually build up its scale, resulting in 170,000 derivative models released by developers worldwide, with a total download volume of over 600 million. Its performance metrics have gradually caught up, making it a benchmark model widely adopted by the enterprise market and Silicon Valley technology companies.
By open-sourcing its software, Alibaba's Qianwen (千问) has become the true embodiment of OpenAI in the public's mind. Now, Alibaba wants to replicate this approach for ordinary users, launching a new challenge against its old rival—OpenAI's ChatGPT (聊GPT). On November 17th, the Qianwen App began its full public beta test.
This "flanking maneuver" tactic, which relies on the open-source ecosystem and moves from the periphery to the core, will inevitably target ordinary consumers—a much larger market with a wider range of scenarios—after conquering the enterprise and developer markets.
Unlike before, now that the model's capabilities have been validated by 1 million B-end clients and the technological trust has been established by millions of developers worldwide, launching an app for ordinary users no longer requires building awareness from scratch. Essentially, it's a well-refined model capability reaching individual users through a more direct interface.
From a timeline perspective, this represents a natural extension of Alibaba's Qwen open-source ecosystem from enterprises and developers to individual users. While Qwen had weathered the challenges of Double 11, businesses like Fliggy, Gaode Maps, and DingTalk had already successfully implemented AI scenarios. The Qianwen App served as the user touchpoint for the entire model family on the consumer side.
The remaining question is whether this ecosystem strategy can win over ordinary users again.
Users are more concerned about the model's "aha moment".
This week, almost everyone was eagerly awaiting the release of the Gemini 3. What people really looked forward to wasn't the product launch itself, but rather a substantial leap in model performance. Google's situation perfectly illustrates this point. Despite previous skepticism, Google has reclaimed its place among the top competitors thanks to breakthroughs in the Gemini 2.5 Pro and Nano Banana models. In today's rapidly evolving technological landscape, temporary leadership is of little significance.
Another noteworthy change in the AI era is that when model performance is powerful enough, products can experience explosive growth even without product launches, relying solely on technical blogs, user word-of-mouth, and improvements in the model's capabilities. Gemini is the best example. — Since April of this year, its app usage has surged by 45%, and its monthly active users have reached 400 million.
This reflects the fundamental law of AI competition: AI products have moved away from the logic of mobile internet. When models and products are strongly bound together, users no longer pay for the UI, but for the "aha moment" of the model.
When there is a technological gap, the product experience merely amplifies it; only when the technological gap is eliminated will competition truly return to the product level.
Over the past three years, Alibaba's self-developed Qwen series models have established a significant leading position in the global open-source AI field. As one of the world's top-performing and widely used open-source models, Qwen has performed exceptionally well in numerous international benchmark tests, continuously driving the evolution of the global AI technology ecosystem through its open-source strategy. From Qwen1 to Qwen3, Alibaba has maintained a high level of R&D investment through high-frequency iterations of the five generations of Qwen models, building a full-modality, full-size model system covering parameter scales from 0.5B to 480B, forming one of the most complete open-source model families in the world.
To date, the Qwen series of models has been downloaded over 600 million times in major global model communities, with more than 170,000 derivative models, making it one of the most widely adopted open-source models by developers and enterprises worldwide. From Silicon Valley tech companies to internationally renowned enterprises, including NVIDIA, Amazon, and Airbnb, all are developing next-generation AI technologies, models, or applications based on Qwen. Stanford University's "Artificial Intelligence Index 2025" report indicates that Alibaba's Qwen ranks third globally in terms of contributions to important models, and the technological gap between China and the US in large-scale models has narrowed to a very small range.
While Qwen has established a strong influence among enterprises and developers, its recognition among ordinary users remains relatively limited. Most people are familiar with ChatGPT and Gemini, but may not realize that the underlying models supporting numerous international AI applications originate from Alibaba. Its technical capabilities have been fully validated, and its ecosystem is maturing—what it currently lacks is a strategic entry point directly accessible to a broad user base.
The release of the Qianwen App completes this crucial step.
This is not an application trial starting from scratch, but a strategic extension based on a mature open-source model. Relying on the solid technical foundation of the Qwen series of models, the Qianwen App possesses the core strength to compete with top international AI products. It is not only an important deployment for Alibaba to enter the AI-to-C market, but also an inevitable step for Qwen to move from the technological foundation to the user's front end.
Users do not need to understand the model architecture; they can experience the powerful open-source model's capabilities for free through the Qwen App: generating professional reports in seconds and automatically converting them into PPTs, accurately identifying products in images and directly linking them to purchase, assisting in processing multiple documents to extract key information, interpreting medical examination reports and providing professional advice. The realization of these functions all stems from Qwen's technological advantages accumulated in global scenarios.
The Qianwen App's mission is to deliver globally validated model capabilities to end users in an efficient and natural way. It is not only a dialogue tool, but also a productivity assistant that can actually solve problems, and an intelligent hub connecting cutting-edge AI technology with real-world scenarios.
Against this backdrop, the Qianwen App is a crucial step for Qwen in moving from "technological leadership" to "user adoption." As top-tier open-source models begin to be offered directly and freely to a wide range of users, the competition in AI technology is entering a more pragmatic new phase.
A Thousand Questions app, utilizing the entire Alibaba ecosystem.
Looking further, Alibaba's AI ambitions are even greater. The launch of the Qianwen App is far more than just the simple implementation of the top-tier Qwen model; it represents Alibaba's strategic move to integrate and reshape its vast ecosystem capabilities and distribute them centrally to every user.
We tested the latest version of the Qianwen App and found that it has integrated Alibaba's ecosystem advantages. For example, when you take a picture and ask it, "What brand is this cup?" it will not only answer your question, but also directly provide links and prices for Taobao, 1688, and Xianyu platforms below, which you can then click to purchase.
Secondly, in terms of content, Qianwen App demonstrates a broader coverage and deeper business integration capabilities. When users seek advice on "a trench coat suitable for autumn," it can not only integrate KOL reviews for intelligent recommendations, but also activate a real-time price comparison system across the entire platform—from official Tmall flagship stores with guaranteed authenticity to outlet channels offering discounts, and even unique designer boutiques, all are readily available.
This is not simply information aggregation, but a deep integration of product, price, and channel data within the Alibaba ecosystem. Qianwen App leverages this to upgrade traditional "information query" into a one-stop "consumer decision-making" process, creating a completely new workflow for users from inspiration to price comparison and order placement. While other AI assistants are still providing suggestions, Qianwen can directly translate choices into actionable steps.
Those who have actually used the Qianwen App will find that its core emphasis on "getting things done" is not unfounded. Its confidence stems from the mature business loop already built within the Alibaba ecosystem: from a world-leading e-commerce platform to local life services covering clothing, food, housing, and transportation, from intelligent and efficient collaborative office tools to accurate and reliable navigation and travel systems, all of these provide a real and rich operating environment for Qianwen to understand user intentions and execute complex tasks.
In actual testing, this ecosystem-level capability is already clearly visible: using the Qianwen App to take a picture and identify objects can accurately identify them and directly redirect to Taobao product links; through the AI "Ask Me" built into Fliggy, users can plan a complete itinerary and recommend flight and hotel packages based on their needs. Such capabilities can be seamlessly integrated into the Qianwen App in the future; and this is just the beginning, with core Alibaba ecosystem applications such as Gaode, Ele.me, DingTalk, and AliHealth being integrated one after another.
This means that users will no longer need to switch between ten apps repeatedly in the future. Simply tell Qianwen, "Help me arrange my business trip to Beijing next week," and it will automatically use Fliggy to check flights, book partner hotels, plan the route from the airport to the venue through Gaode Maps, generate meeting minutes and invitations on DingTalk, and even compare prices and buy necessary travel supplies for you directly on Taobao.
This is no longer a simple aggregation of functions, but a systematic manifestation of ecosystem-level agent capabilities. No matter how powerful ChatGPT and Gemini models are, they cannot orchestrate a complete business ecosystem covering e-commerce, local services, office work, and travel. When all these capabilities are integrated into the Qianwen App, it is no longer a simple AI dialogue tool, but has evolved into a super AI portal for the entire Alibaba ecosystem facing consumers.
The technological foundation has been thoroughly validated, user touchpoints have been successfully established through the Qianwen App, and true ecosystem integration is in full swing. This is the real potential that differentiates the Qianwen App from its competitors.