In September 2025, Alibaba's core decision-makers and Quark and other business executives held a series of meetings to "affect Alibaba's future destiny".

In this series of meetings, the group's core decision-making level made a decision: to fully promote the C-end AI project called "Qianwen APP", and regard "Qianwen APP" as the core of Alibaba's entire C-end AI market layout: Qianwen APP will directly enter the C-end market; In addition, the AI C-end products that Alibaba has released and other AI C-end products that will be released in the future will also revolve around the "Qianwen APP" and the "Tongyi Qianwen" large model behind it.

This means that Quark network disks, browsers, upcoming AI glasses, and other AI C-end scenarios that Alibaba will try will all be included in the big ecosystem of Qianwen. In other words, Qianwen's hope is to "become Taotian's C-end mental symbol for Alibaba e-commerce".

At this series of meetings, some high-level executives described the "Qianwen APP" as "a key battle affecting Alibaba's future". According to people familiar with the matter, a very important senior figure of the group made it clear that the "Qianwen APP" is not only a strategic project that affects the future fate of Alibaba, but also an "overall battle" that requires cooperation and joint efforts of all departments of the group. After that, on a Wednesday at the end of October, a kick-off meeting was organized with the participation of internal core cadres.

96 hours after the kick-off meeting is Monday.

In these 96 hours, more than 500 engineers mobilized from Beijing, Guangzhou and Hangzhou have secretly moved into the third and fourth floors of the C4 floor of Hangzhou Ali Xixi Park. They were almost the "elite generation" of Alibaba's internal AI products at that time: there were post-95 geniuses who had been carefully trained by Alibaba for many years since graduation, AI academic experts who had been excavated from overseas with a lot of money from Alibaba, senior product managers who had participated in many products and fought many tough battles, and experts with expertise in multi-modal and other subdivisions......

The workstations, accommodation, and project division of more than 500 people were quickly "done" within 96 hours, and the expedited printed special work cards were issued to them after they moved into the C4 secret work area on Monday. Most of them, until they got on the plane to Hangzhou, did not know that they would be committed to a "top secret" strategic project within Alibaba, and they were only told that the company urgently needed them to participate in a key project.

In the following week, the number of elites parachuted in increased to nearly 900, which became a new record for Alibaba to transfer the number of elites from different departments for a single project.

At the end of the fourth floor of the C4 floor, there is a room named "War Room" - it is actually a large conference room. Wu Jia, the person in charge of Qianwen APP, moved his office here, and on the side of the war room, it surrounded the workstations of Wu Jia's core subordinates and the heads of the core section of Qianwen APP.

Here, it has become the "battlefield command post" of Ali's tough battle.

There are no OKRs, no rules, and no master planning blueprint called "roadmap." More than 900 people are closed on the third and fourth floors of the C4 building in the form of a project team, and the data reports and user analysis generated by manual and AI will be sent to each "frontline combatant" every night. Every week, the weekly goals set by the "war room" meeting will be refined into a specific work plan and split into each member of the project team.

Alibaba is rare, in the form of a "project team" on the battlefield, rather than a large organizational structure, to promote the "birth" of Qianwen.

More than a month later, on November 17, the Qianwen APP was officially launched. Seven days later, it was downloaded over 10 million. Tiger Sniff learned that this data greatly exceeded the "expectations" previously set by the group's core level for the team. 10 days after Qianwen was launched, Quark AI glasses were officially launched as a hardware part of the "Qianwen Ecosystem", and Tiger Sniff learned that the total number of orders for Quark AI glasses so far has been several times higher than the previous team's expectations.

Judging from the data, the first two battles of Qianwen APP and Quark AI Glasses ended in small victories. But it is undeniable that a bigger tough battle has been put in front of Ali: from the perspective of C-end AI products, Doubao, DeepSeek, and Yuanbao firmly occupy the top three in the market, and behind them, Kimi and other six tigers are also struggling to survive; In the AI glasses market, Meta, which has a first-mover advantage, occupies a large share of the global market, and brands such as Rokid, Xiaomi, and Yingmu have already won a share that is enough to consolidate mental cognition in the domestic AI glasses market.

It is worth noting that whether it is in the C-end AI APP or AI glasses, these opponents in front of Ali are extremely strong. Taking Doubao as an example, through the traffic endowment of "different from ordinary people", Doubao has risen from the era of "AI Six Little Tigers Dominating the World" in 2024, and in the face of the successive impacts of DeepSeek and Yuanbao, it has held on to the country and become a rare "100 million users" AI single product in China. In the field of AI glasses, Rokid has withstood the impact of giants in the industry as an entrepreneur, and won the "mythical record" of "5-day sales of 40,000 units" in the third quarter of this year.

Qianwen is regarded as the third key fate battle of Ali in the Wu Yongming era. One is AI+ cloud, one is Taobao flash sale, and the other is Qianwen. A relevant person told Tiger Sniff, "One of the internal consensus is that if it succeeds in the AI C-end market, then Alibaba's imagination may be several times larger than it is today." ”

Fateful decisions that took place in the summer

Tiger Sniff exclusively learned that before the summer of 2025, the "Qianwen Plan" has not become the first choice for Alibaba's core decision-making level.

In fact, until the beginning of this year, Quark was still the protagonist of Alibaba AI's C-end narrative.

A key factor has become a bargaining chip that affects the direction of the decision-making scale.

In September 2025, Alibaba Cloud released Qwen3-Max, the latest flagship model with a parameter scale of more than one trillion yuan. Tiger Sniff learned that when Alibaba's core decision-making layer experienced Qwen3-Max's powerful multi-modal capabilities and reasoning research capabilities, the decision-making level's confidence in "doing a pure C-end AI application" has increased significantly.

A relevant person told Tiger Sniff that there is a core decision-making layer familiar with technology and AI, and after personally trying out most of the C-end AI model tools on the market, he compared Qwen3-Max, and the person believes that Alibaba should adjust its previous thinking as soon as possible and directly enter the C-end market. "We may not have a traffic advantage or a social advantage, but our biggest confidence is model capabilities. We found that a large number of domestic head AI applications have actually fine-tuned the Qwen model, so why don't we do it ourselves? ”

This triggered a series of "decision-making meetings" at the beginning of the article.

Judging from the information learned by Tiger Sniff, the emergence of Qianwen APP is not an established strategy many years ago, but a dynamic decision made based on the latest internal and external situation.

Once the general direction is determined, it is time to decide what to do.

In the third quarter of 2025, Alibaba's core decision-making level and business held several meetings together to discuss how to enter the C-end AI market.

After many communications, the core layer gradually formed three basic consensuses, which also became the "longzhong pair" of the entire Alibaba AI to C:

Qianwen APP is the core, but it is not just an APP, it is necessary to let Qianwen cut into more scenarios (for example, Qianwen APP is integrated into Quark AI glasses in the form of AI assistants, and also into browser products in a similar form)

2. It is not a simple chat companion tool, but meets the needs of users to chat while providing "efficiency and professionalism" and becoming a productivity tool for users. Considering the needs of today's users for chat, Alibaba's model capabilities are also exerted. Based on this, the core layer gave a six-character policy of "can chat, can do things".

3. Now (referring to September 2025) is a critical moment, because the Qwen3-Max model is capable of doing this, and we must seize this historical window period to make an AI application that can "put the entire Alibaba ecosystem into it". In this historic window period, investment must be resolute and cannot be hesitated

After the emergence of these three basic consensuses, the Qianwen APP project can be described as "unimpeded" within Alibaba: the most elite genius engineers, the most flexible project and management methods, and high resource investment have followed.

A question that arises is, how exactly should the boundary between quarks and thousand questions be defined?

Relevant people told Tiger Sniff that based on three basic consensuses, the team has made a new positioning for Quark: Quark no longer plays the "core" of Alibaba's AI to C matter, but instead focuses on what it is best at "search and post-search business (such as network disk and scanning)".

It is worth noting that from the perspective of organizational relationship, Quark and Qianwen APP are both affiliated to the "Alibaba Intelligent Information Business Group", and Wu Jia, as the president of this business group, also serves as the CEO of Quark and the head of Qianwen APP. In the process of promoting the Qianwen APP project, a large number of technical capabilities derived from Quark have been embedded in the Qianwen APP. One of the key responsibilities of Wu Jia is to allocate product capabilities and resources based on the different positioning of the two products.

A "general mobilization" of Alibaba

"War Room" on the fourth floor of the C4 floor of Alixixi Park, Hangzhou.

The house is full of drinks and snacks, and here Wu Jia and the core members of Qianwen have endless review meetings and seminars every day, and they often don't care about eating and grab a bag of snacks to satisfy their stomachs "on the spot".

The Qianwen project team is a "special zone" within Alibaba.

One detail is enough to show the specialness of this project team, it is separated from the traditional Alibaba's organizational model, management style, and product birth system, and it is a product born on the "front line of war". The highest level of the group has given this group of "younger commanders" at the highest level of the front line great flexibility and space.

Tiger Sniff learned that in order to ensure the creativity of the Qianwen project team, the highest level of the group did not let the team adopt OKR and other management methods. Most of the more than 900 engineers who came to the space did not have specific OKRs for the Qianwen project in the past two months.

Tactical goals are updated in a timely manner based on real data and user feedback, and with the process of product iteration, everyone will clearly understand the "specific output goals for each week".

Equally flexible is the budget system. Unlike many large projects within Alibaba, the cost application of the Qianwen project is based on the unit of the week, according to the "key nodes, key outputs, and key weekly goals" of the next week, and the approval and evaluation process of the entire budget is very fast.

The product itself also adopts a strategy of "Alibaba's overall war": the different AI capabilities embedded in the Qianwen APP actually come from the capabilities previously accumulated by different business lines.

For example, multi-modal capabilities are completed by the elite parachuted from teams such as Tongyi Large Model; Test paper recognition and AI medical-related capabilities are undertaken by Quark's elite talents who are good at these aspects.

For each competency section, subdivided sub-project groups have been established, and students belonging to these groups may have been distributed in different business sectors before. When a capability (feature) is completed and goes live for testing, a separate data analysis team begins daily data tracking. If the user data (user penetration, user feedback) of a certain capability is not as expected, the data analysis team will meet with the sub-project team to discuss the logic behind it, and then quickly iterate back.

DAU等

A series of data (DAU, etc.) at the user level has become the key "goal" of Qianwen's entire project team, but not all.

Tiger Sniff learned that because in the "three basic consensus", the Qianwen project is regarded as "not an AI that simply accompanies chatting" but "a productivity tool AI that can meet chat needs and improve user efficiency and professionalism", so when considering the user dimension, the team considers "quantity" and "quality". There is a separate group that will conduct in-depth research on the characteristics and needs of all Qianwen APP users, and study how to satisfy users in terms of chat companionship and productivity.

In terms of goals and considerations, the group's decision-makers have also given Qianwen APP a different "tolerance space" than before. Tiger Sniff learned that until the day of the launch of the Qianwen APP, the decision-making level did not give the project team a clear "red line of data indicators (that is, it cannot be less than how much)", but the members of the decision-making level basically became the seed users of the Qianwen APP, and they would often send their experience feedback or suggestions to the group in some separate communication groups.

From these perspectives, Qianwen APP may bring Alibaba not only an AI C-end product, it actually represents a new "collaboration method" within Alibaba. The author can give an example: these elites who have parachuted into Qianwen actually have their own clear OKRs in their own business sectors before receiving the "transfer notice". After joining the Qianwen project team, their existing OKRs will be in a state of "pending", and Tiger Sniff learned that Alibaba has quickly formed a set of guarantee plans for the performance and treatment of these elites in the past few months. And these institutional models actually create a paradigm for similar "super projects" within Alibaba in the future.

In terms of collaboration, hundreds of elite projects like Qianwen are not every day within Alibaba. Moreover, as mentioned above, Qianwen actually "adopts various AI capabilities from different departments of Alibaba." This kind of sincere cooperation across departments to "bring out the best talents and the most precious technical secrets" can only appear under the premise of Alibaba's top management promoting the "overall war". But obviously, this also serves as a demonstration for Alibaba's in-depth cross-departmental collaboration in the future.

Conclusion: It's just begun, it's a tough battle

Looking at it calmly, the Alibaba AI to C war triggered by the Qianwen APP has actually just fired the first shot.

For Alibaba, there is never a moment to celebrate.

Qianwen APP and Quark AI glasses are essentially competing for the C-end AI entrance, and the C-end AI entrance is a battleground for Byte, Tencent, Baidu, and Meituan.

When communicating with a number of people related to Alibaba's AI to C products, the author asked a key question: How much do you invest in traffic?

Judging from the information perceived by the author, they are very aware of the "traffic magic" behind the rise of Doubao, DeepSeek, and Yuanbao in the past two years; They also know that it is difficult for Alibaba to "wave blank checks and invite guests and friends" on the traffic side like some friends.

A relevant person told Tiger Sniff that Alibaba is very calm and restrained in AI to C traffic investment.

"Internally, we will soberly believe that this is not the same thing as Taobao flash sales, and we cannot treat Qianwen like we treat instant retail business." The person believes that the internal situation is very sober, instant retail is for the entire C-end market, and it embraces every C-end user; But Qianwen's positioning is not an entertainment product, not a simple chat product. The above-mentioned person said.

But this has triggered the author's thinking: judging from the fate of several popular C-end AI products in China in the past two years, they often need to seize the critical window period to quickly achieve "user fission" at the beginning of their adoption, and then quickly cross the 10 million monthly active mark, and finally evolve into a national-level (100 million users) explosive product.

The hidden difficulty is that whether it is bean bags or ingots, they are still in the "strategy cycle" to this day, and they are still eager to continue to expand; And in the coming months, DeepSeek is likely to release its long-awaited R2. If we consider that many of the above players are "big pools" with traffic itself, will Alibaba's expansion in the C-end market fall into the plot of what was once quark: product capabilities are praised, but marketing and momentum are interpreted as "slightly inferior"?

It doesn't stop there.

According to the information learned by Tiger Sniff, based on Qianwen, Alibaba tried to integrate the entire ecosystem, including e-commerce. This is obviously a huge space for imagination. But to this day, including OpenAI, the circle is still in the exploration stage of what AI e-commerce looks like. The reshaping of e-commerce by AI itself is divided into two sides:

If you have an AI super entrance, then how to attract traffic to e-commerce based on this entrance is what Qianwen has to do

How do you use AI to reconstruct existing e-commerce businesses, which is what Taotian is doing

The challenge for Alibaba is that these two things need to be promoted simultaneously.

This is destined to be a tough battle.