Business judgment on AI products
There is public usage data. The useful view is what it does, its scale, and whether the change lasts.
01
Start inside the user's day · Public facts + observable behavior · 2026-08-28
Overseas users open a page to talk to 千问, download models, and read API docs — the international front door.
Public materials do not yet show how users complete this job today or what they replace.
The product targets friction in this job, but public user evidence does not yet show the cost, frequency, or consequence of leaving it unsolved.
One model, several doors: domestic for scale, international for reach, open-source to become the default. The trend is open models surviving by being built into other people's stacks. The entry is overseas developers and chat users. Chat is free; money is usage on the cloud.
A public record shows 35.43M monthly visits and -2.85% month-over-month growth. That explains the attention, but product-level retention and payment are not yet verified.
① Whether the 35.43M international-site visits return to positive MoM — organic growth abroad after the holiday dividend fades; ② Retention of the Qianwen app after subsidies taper — how much of 203M was bought; ③ Whether the open-model release cadence changes after Lin Junyang's departure — the nex…
Worth dissecting. A public record shows 35.43M monthly visits and -2.85% month-over-month growth. That explains the attention, but product-level retention and payment are not yet verified.
one model base, three entries — domestic app for scale, international site for reach, open source for ecosystem — each feeds the others, and growth is never bet on a single door. For platform products, "one base, multiple entries, each with a job" is steadier than "one entry done perfectly."
Consumer chat is free. Monetization runs through Alibaba Cloud (Qwen models metered on the Bailian platform and international cloud), plus the cloud compute consumed by the open ecosystem. Alibaba does not disclose Qwen-specific revenue. ① Whether the 35.43M international-site visits return to positive MoM — organic growth abroad after the holiday dividend fades; ② Retention of the Qianwen app after subsidies taper — how much of 203M was bought; ③ Whether the open-model release cadence changes after Lin Junyang's departure — the nex…
Overseas users open a page to talk to 千问, download models, and read API docs — the international front door.
A public record shows 35.43M monthly visits and -2.85% month-over-month growth. That explains the attention, but product-level retention and payment are not yet verified.
An English validation note will follow from the public evidence.
The product claims to help users complete: “Overseas users open a page to talk to 千问, download models, and read API docs — the international fro”. User evidence has not yet verified pain intensity or the cost of doing without it.
The assessment is recorded; an English explanation is pending.
The assessment is recorded; an English explanation is pending.
The assessment is recorded; an English explanation is pending.
02
Market comparison
The Chinese–English market comparison is not complete yet. A conclusion follows only after its coverage and verifiable evidence are recorded.
03
The call and next move come first; the full read retains the evidence and counterevidence.
Alibaba Qwen's international portal: chat, the model library, and API docs in one place — the front door for overseas users to Qwen, and the official site of the most-used open-source model family in the world.
Alibaba's Tongyi Lab. Qwen was led from late 2022 by Lin Junyang (formerly a Tongyi Lab scientist) until his departure in early 2026. This is Alibaba's international entry point, sharing one model family with the domestic Qianwen app across two sites.
Note on entries: Alibaba's domestic Qianwen app (star of the Spring Festival campaign, 203M MAU in February 2026) and this Qwen.ai international site (the 35.43M monthly visits traffic board tracks) are different things — the former is a Chinese super-app entry, the latter an overseas portal; a layer further out sits the open-model ecosystem (200K+ derivatives on Hugging Face).
Read: Qwen's strength is three entry layers sharing one model base — the domestic app for scale, the international site for reach, open models for the ecosystem. Any single layer looks incomplete alone; together they make the "world's most-used open model" story.
Two kinds of people, two replaced behaviors.
For overseas end users: the habit of doing everything inside one closed chat tool. A free, open model with top-three capability became a mainstream alternative to ChatGPT — and the door anyone researching open models has to open.
For developers: the default of building open applications on Llama. Previously, Meta's Llama was the default for self-hosting an open model; Qwen replaced that default with download volume and derivative-model counts. That is not winning a leaderboard; it is winning a default behavior.
Read: ecosystem replacement is the hardest to reverse — switching models means retraining, retuning, and re-releasing; once a team's stack sits on Qwen, the migration cost itself becomes the retention moat.
Consumer chat is free. Monetization runs through Alibaba Cloud (Qwen models metered on the Bailian platform and international cloud), plus the cloud compute consumed by the open ecosystem. Alibaba does not disclose Qwen-specific revenue.
| Dimension | Call |
|---|---|
| Founder-product fit | Lin Junyang turned Qwen into the world's #1 open model in three years — proven by results, though he has since left |
| Product insight | Open source plus three layered entries; uses the ecosystem to replace a default behavior — a clear strategy |
| Execution quality | Qwen3-Max in the global top three, first in open derivatives — engineering and ecosystem both strong |
| Timing | The domestic super-app entry is already decided by subsidy wars; the international site is a slow variable; Qwen is already at the table |
Worth watching. This is a "verified" stable leader: 203M MAU, +552%, #1 in open source — the data is real. The international site's 35.43M visits at -2.85% MoM is the normal cool-down after the Spring Festival dividend, not a problem — because Qwen's growth story was never about this one chat entry.
Read: the real thing to watch is whether the three-layer structure feeds itself — open source feeds developers, developers bring cloud revenue, and the domestic app holds scale. In that structure the international site is the brand's face, not the whole business. The things to watch are continuity of direction after Lin's departure, and domestic retention once subsidies taper off.
① Whether the 35.43M international-site visits return to positive MoM — organic growth abroad after the holiday dividend fades ② Retention of the Qianwen app after subsidies taper — how much of 203M was bought ③ Whether the open-model release cadence changes after Lin Junyang's departure — the next model generation is the test of strategic continuity
Product logic: one model base, three entries — domestic app for scale, international site for reach, open source for ecosystem — each feeds the others, and growth is never bet on a single door. For platform products, "one base, multiple entries, each with a job" is steadier than "one entry done perfectly."
Open-source strategy: the goal of open source is not reputation, it is replacing a default behavior — get teams to stack on your model, and migration cost becomes your moat. Judge open source by derivative-model count and team adoption, not by download leaderboards.
Worth watching. Real data, settled position, and a working three-layer structure. Track retention after subsidies taper and the continuity of the open-source roadmap, not the 2.85% MoM on the international site.
05
Use these searches when the official site is missing or the current link is only a lead.