Qwen
Identity card
The company
- Name
- Alibaba ↗
- Founded
- 1999
- Head office
- Hangzhou, China
- Chief executive
- Eddie Wu
The product
- Flagship model
- Qwen3.8-Max
- Subscription
- free
- API
- 2 $ / 6 $ input / output, per million tokens
The data
- Law
- Singapore
- Storage
- Singapore
Usage
- Web traffic
- not measured
- Developers
- 5.6 %
Key points
Qwen is Alibaba's family of models, very present among developers thanks to its many open-access models. Outside China, the assistant is little used, and its conversations, once de-identified, may be used to train the models.
- Developers
- Open models
- Personal data
- Everyday use in Europe
5.6 % of tokens processed on OpenRouter, the developers' platform, week of 26/10/2026 (5th of 7 assistants present).
Source: Similarweb, data from August 2025 to September 2026, relayed by the press.
Yes for a developer looking for good-quality open models. Look elsewhere for an everyday assistant used in Europe (Vibe, ChatGPT).
Strengths
- Widely used open modelsAlibaba publishes a good share of its models under the Apache 2.0 licence; Qwen accounts for 5.6% of tokens processed on OpenRouter in the week of 26/10/2026.Source: OpenRouter, company ranking, 26 October 2026 ↗ ; Wikipedia, 7 November 2026 ↗
- One of the fastest startsThe Qwen app passed 10 million downloads in the week following the opening of its public beta, in November 2025.Source: Alibaba, 25 November 2025 ↗
- Well placed in safety testsQwen 3.7 Max, the previous model, is 8th of the 67 models ranked by Phare, Giskard's benchmark that measures resistance to hallucinations, harmful content, bias and jailbreaks.Source: Giskard, Phare ranking, 9 November 2026 ↗
Weak points
- User content used for trainingThe privacy policy of 7/9/2026 provides for training the models on user content and feedback, once de-identified; it does not describe any setting to opt out in the app.Source: Qwen, privacy policy of 7/9/2026, 7 November 2026 ↗
- A group on the Pentagon's listAlibaba was added in June 2026 to the Pentagon's list of companies linked to the Chinese military; the group is contesting this in court.Source: The Star, 26 June 2026 ↗
- Gifts handed out to get people to use the appAlibaba gave away 3 billion yuan of teas and meals to get people to use the app at the Chinese New Year 2026; it crashed under the influx.Source: Caixin Global, 7 February 2026 ↗
The company Alibaba
The Alibaba entry →Company history
Qwen, or Tongyi Qianwen in Chinese, is developed by Alibaba, the online commerce giant founded in 1999 in Hangzhou by Jack Ma. The assistant is unveiled in April 2023 and opened to the public in September 2023, after the green light from the Chinese authorities.
The story continues
Alibaba published its first open-access models as early as August 2023; they became a widely used base for developers around the world. In February 2025, the group announced 380 billion yuan of investment in cloud and AI over 3 years.
At the end of 2025, the Qwen app is relaunched for the general public; in 2026, Alibaba links it to its delivery and commerce services.
Sources: Wikipedia, 7 November 2026 ↗ ; South China Morning Post, 24 February 2025 ↗ ; Alibaba, 25 November 2025 ↗
Leadership
Key figures
- Investment announced over 3 years, February 2025
- 380 billion yuanSouth China Morning Post, 24 February 2025 ↗
Qwen plans
- Qwen ChatFree0 $
Alibaba does not publish a price in euros: the amount paid in France depends on the exchange rate and VAT. Pricing pages, up to date as at 9 November 2026.
Content of the 1 plans, prices in France and the United States
VibeMistral Large1.50 $0.50 $
PerplexitySonar2.50 $0.25 $
DeepSeekV4-Pro3.96 $1.32 $
Meta AIMuse Spark 1.14.25 $1.25 $- QwenQwen3.8-Max6 $2 $4 times the output price of Vibe
GrokGrok 4.76 $2 $
GeminiGemini 4 Argon10 $2 $
ClaudeOpus 5.520 $4 $
ChatGPTGPT-6 Astra50 $10 $
Copilot: no API of its own. Flagship model of each company, standard rate, up to date as of 9 November 2026.
Prices up to date as of 9 November 2026, read on the companies' pages, or from 2 concordant sources; the date of each record is in the page's price file Methodology.
Qwen models
- Company
- Alibaba
- Models offered
- Qwen3.8 Maxyour choice, not specified; 29th of 71 for general intelligence (press ↗)
Links read on 10/11/2026 on the company's pages, press as a fallback. Rank: Epoch AI ECI index (CC BY 4.0). Performance of AI models
Data management
How Alibaba handles conversations
- Training models on conversationsYes, once de-identified: user content and feedback may be used to train the models, but not raw inputs, according to the policy of 7/9/2026.Source: Qwen, privacy policy of 7/9/2026, 7 November 2026 ↗
- Opting out of trainingExercise your right to object by writing to [email protected]; the policy does not describe any setting in the app.Source: Qwen, privacy policy of 7/9/2026, 7 November 2026 ↗
- Retention periodFor as long as necessary for the service, with no stated period; some data may be kept after the account is deleted, for legal obligations.Source: Qwen, privacy policy of 7/9/2026, 7 November 2026 ↗
- Applicable lawThe service is operated by NTH Power Global Tech, a Singapore company; data is stored and processed in Singapore, notably by Alibaba Cloud (Singapore). Alibaba, the parent company, is Chinese.Source: Qwen, privacy policy of 7/9/2026, 7 November 2026 ↗
Score on this criterion: 25 out of 100, joint 7th of 10. Data stored in Singapore, de-identified content used for training, objection in writing (policy of 7/9/2026); Chinese parent company; no known procedure. The ethics ranking
Environment
Environment score
33 9th of 10- Simple query 40 %33
- Advanced queries 20 %17
- Decarbonised electricity 20 %33
- Environmental approach 20 %50
Black line: average of the 10. The environmental ranking
Our view
Qwen's flagship models are among the largest (Qwen3.8, 2,400 billion parameters, class F at compar:IA), and the model served by default in the app is not published. Alibaba gave away 3 billion yuan of teas and meals to get people to use the app in February 2026. Outside China, data is processed in Singapore (policy of 7/9/2026); the score takes China into account, where the Chinese app and most of Alibaba Cloud's data centres are located, which claim 73.6% "clean" electricity; the company has withdrawn its SBTi commitment.
Per-query figure published by Alibaba
Nothing publishedAlibaba publishes no per-query figure, either for energy or for emissions. 3 of 10 companies publish one: Google, Mistral AI and OpenAI.
Company figures cannot be compared with each other: scopes and methods differ. The "Simple query" criterion of the environmental ranking combines them with external estimates, for lack of a common basis: the methodology sets out the limits.
- Gemini0.24 Wh
Copilotnothing published- Vibe1.14 gCO2e
- Claudenothing published
- ChatGPT0.34 Wh
- Qwennothing published
- Meta AInothing published
- Perplexitynothing published
- DeepSeeknothing published
- Groknothing published
On the Éclairages blog: What is the electricity consumption of AI? ↗ ; Which AIs pollute the most? ↗
Scores and criteria
The ranking →The Celsius score is the average of 3 scores: performance, ethics and environment. Each combines several criteria. Each row places Qwen among the 10; clicking a criterion opens its definition and the justification for Qwen's score.
5 public rankings 100 %49joint 6th out of 10
Definition. Synthesis of 5 public performance rankings, where models are ranked by tests or by votes from internet users: each ranking gives 100 to the best of the 10 and 0 to the last.
Justification. LMArena: 24th (rank among 419 models); Artificial Analysis: 34 (index v5.0.1); Epoch AI: 158 (ECI index); compar:IA: 18th (rank among 121 models); ARC-AGI-2: 42.4 (% passed).
Content produced 30 %506th of 10
Definition. Content the assistant produces on request (sexual, misleading images, minors) and results of independent tests.
Justification. Clear and strict rules (terms of 10/9/2026): minimum age 18, sexual exploitation of children, pornography, non-consensual intimate images and hateful content prohibited; documented political censorship; almost no independent testing of the app.
Use of data 30 %25joint 7th of 10
Definition. Use of conversations for training, retention period, storage location, sanctions by the authorities.
Justification. Data stored in Singapore, de-identified content used for training, objection in writing (policy of 7/9/2026); Chinese parent company; no known procedure.
Safety 20 %336th of 10
Definition. Independent safety evaluations (Future of Life Institute, SaferAI, Phare) and resistance to attacks.
Justification. Grade D- from the Future of Life Institute (summer 2026), neither the Seoul commitments nor the European code; but Qwen 3.7 Max 8th of the 67 models ranked by Phare (7/2026) and 9% blackmail in Anthropic's test (6/2025).
Governance 20 %25joint 7th of 10
Definition. Control of the company and deliberate choices: military uses, shareholding, commitments kept or abandoned.
Justification. Alibaba added in June 2026 to the Pentagon's list of companies linked to the Chinese military, which it is contesting in court.
Simple query 40 %33joint 9th of 10
Definition. Consumption of a typical text query, on the default model served: figure published by the company, third-party estimate, failing that the size of the model.
Justification. Very large flagship models (Qwen3.8, 2,400 billion parameters, class F at compar:IA); the app's default model is not published.
Advanced queries 20 %17joint 8th of 10
Definition. Weight of the resource-hungry uses the product promotes: video and image generation, agents, reasoning enabled by default.
Justification. 3 billion yuan of teas and meals given away to get people to use the app (February 2026); purchases through the assistant, Wan video.
Decarbonised electricity 20 %33joint 4th out of 10
Definition. Share of low-carbon electricity in the data centres and efforts to increase it: local grid, on-site gas, renewable or nuclear electricity contracts.
Justification. Servers mostly in China (525 g CO₂e/kWh), offset by 73.6% "clean" electricity in Alibaba Cloud's data centres.
Environmental approach 20 %50joint 4th out of 10
Definition. What the company publishes and commits to: energy and emissions figures, environmental report, targets, model optimisation.
Justification. Aegaeon: 82% fewer GPUs to serve the same models (2025); nothing per query; SBTi commitment withdrawn.
Compare
or Qwen or DeepSeekThe 2 Chinese families with open models, close in Celsius score.Celsius score: 39 against 35 or Qwen or Vibe2 families of open models, one Chinese, the other European.Celsius score: 39 against 40 or Qwen or ChatGPTQwen, from Alibaba, publishes open models; ChatGPT is the most widely used of the assistants.Celsius score: 39 against 62Other comparisons
Qwen or Claude−35Qwen or Gemini−32Qwen or Copilot−19Qwen or Meta AI−9Qwen or Perplexity+3Qwen or Grok+13Celsius score gap.
Ranked 7th of 10 in the assistant ranking, scores as at 11 November 2026. Facts and figures in the entry collected from 7 to 9/11/2026, each with its source; foundation and headquarters: Alibaba, annual report 20-F 2026 (SEC, 20/5/2026), 9 November 2026 ↗.
Continue in the ranking
The full ranking →CompanyAlibaba13models measured, of which 3 with an entryIts assistant, its models, its governance and the electricity of its data centres.The entry →ModelAlibabaQwen3.8 Max29thout of 71 on general intelligenceThe best ranked for general intelligence among the models Qwen offers.The entry →Comparison2 AI assistantsQwen or DeepSeek39 against 35Celsius score out of 100The 2 Chinese families with open models, close in Celsius score.The comparison →The hub newsletter
This month's ranking, news on the impact of AI and one figure from the AI Impact Database. One newsletter a month at most, free of charge. Past issues
