Qwen (Tongyi Qianwen) is an AI model family owned by Alibaba, developed by the Qwen Team and Alibaba Cloud, and released publicly since 2023. It is not a separate company. The listed company behind it is Alibaba (NYSE: BABA, HKEX: 9988).
What sets it apart is the “full stack”: Alibaba owns the models, a public cloud (Alibaba Cloud), its own services such as Taobao and Alipay, and its own AI chips. Among Chinese large-model companies, that setup is rare.
Key Data Snapshot
| Item | Data |
|---|---|
| Product name | Tongyi Qianwen / Qwen (model family + Qwen App) |
| Company behind it | Alibaba Group / Alibaba Cloud, developed by the Qwen Team |
| Listing status | Alibaba is listed (NYSE: BABA, HKEX: 9988); Qwen has no separate funding or valuation |
| Founded | Alibaba in 1999; Qwen released publicly since 2023 |
| Headquarters | Hangzhou, China |
| Leadership | Group CEO Eddie Wu (吳泳銘, took over in 2023); CFO Toby Xu |
| Flagship model | Qwen3.8-Max (released August 2026, open weights) |
| License | Varies by model: most small models Apache-2.0; flagship has open weights under a custom license; some models API-only |
| Qwen App monthly active users | About 167 million (QuestMobile, June 2026), No. 2 among China’s AI-native apps |
| In-house chips | T-Head Zhenwu V900 (unveiled September 2026); Alibaba has not said which chips Qwen is trained on |
A Quick Tour of Seven Dimensions
① Technology and product: Released in August 2026, Qwen3.8-Max has about 2.4 trillion parameters. It is the first Max-class Qwen with open weights, breaking the old “closed flagship” approach. But it uses Alibaba’s own custom license, not Apache-2.0: once your user count or revenue passes a certain size, extra terms apply. In September Alibaba launched Qwen3.8-Omni-Flash, an API-only omni-modal model, and said the next generation, Qwen 4, is in training. Most benchmarks are Alibaba’s own tests, so read them as official claims.
② Customers and positioning: On the consumer side there is the Qwen App. QuestMobile counted about 167 million monthly active users in June 2026, second only to Doubao. Businesses and developers pay for the API through Alibaba Cloud Model Studio (Bailian), which connects to Alibaba services such as Taobao and Alipay. Rivals in China include DeepSeek, Doubao, Kimi, and GLM.
③ Ecosystem and partnerships: Open models go on Hugging Face and ModelScope to build a developer community, and business demand is then steered back to Alibaba Cloud. Qwen’s biggest advantage is that it does not fight alone: the whole of Alibaba Cloud stands behind it.
④ Valuation and financials: Qwen has no standalone valuation, so look at Alibaba’s financials. In the quarter ended June 30, 2026, Alibaba’s “AI cloud and compute services” revenue was about RMB 48.4 billion, up 45% year over year; capital spending was about RMB 67.7 billion, up 75% (official results). In March 2026 Alibaba set a target of more than $100 billion in combined annual external cloud and AI revenue within five years. That is a target, not actual revenue.
⑤ Commercialization and regulation: Open weights attract developers but can also eat into Alibaba’s own API income. Price competition among Chinese large models is fierce, so margins are a long-term variable. In China, generative AI services must first be filed with the CAC and meet content-safety rules.
⑥ Geopolitics and supply chain: US export controls on advanced AI chips are the backdrop to Alibaba’s in-house chips. In June 2026 the US Defense Department added Alibaba to its “Chinese military companies” list (1260H). Alibaba sued, calling the listing “without factual or legal basis”; as of October 2026, the court has not ruled on the listing itself. In September 2026 an advisory from the US NSA, FBI, and CISA, and a report from Anthropic, also named Alibaba, saying it trained its models on answers from US models (known in the industry as “distillation”). These are accusations, not court findings.
⑦ Leadership and governance: Qwen has no governance structure of its own; Alibaba holds control. Eddie Wu has led the AI and cloud strategy since becoming CEO in 2023; founder Jack Ma stepped down from executive roles in 2019. Qwen’s model cards are mostly credited to the Qwen Team, without named researchers.
Key Milestones
| Date | Milestone |
|---|---|
| 2023-08 | Qwen completes the CAC generative AI service filing (early versions) |
| 2025-04 | Qwen3 series released and open-sourced |
| 2026-03 | Annual external cloud and AI revenue target: over $100 billion within five years |
| 2026-06 | US Defense Department adds Alibaba to the 1260H list; Alibaba sues |
| 2026-08 | Qwen3.8-Max released; open weights on August 12 (custom license) |
| 2026-09 | Qwen3.8-Omni-Flash launched (API only); Zhenwu V900 chip unveiled at the Apsara Conference, Qwen 4 announced as in training |
This table was last compiled in October 2026.
Further Reading
- Other Chinese models: DeepSeek, Doubao, Kimi, GLM
- The full hardware chain: The AI Hardware Supply Chain, End to End
Penchan’s summary
When you read Qwen news, separate two things. First, Qwen is one of Alibaba’s product lines: valuation and financials come from Alibaba’s results, and a “revenue target” is not booked revenue. Second, since August 2026 even the flagship has open weights, so “open source” can’t be judged as a whole; check each model’s own license.