Shirt a Phobia

Sample Page

Qwen (also known as Tongyi Qianwen, Chinese: 通义千问; pinyin: Tōngyì Qiānwèn) is a family of large language models developed by Alibaba Cloud. Many Qwen models are distributed under the free and open-source Apache 2.0 license, the source-available Qwen License,[1] or the non-commercial Qwen Research License;[2] other proprietary Qwen models are served through Alibaba Cloud.[3]

Models

Qwen

Alibaba launched a beta of Qwen in April 2023 under the name Tongyi Qianwen, then opened it for public use in September 2023 after regulatory clearance.[4][5]

The model’s architecture was based on the Llama architecture developed by Meta AI.[6][7] In December 2023, it released its 72B and 1.8B models for download, while Qwen 7B weights were released in August.[8][9] Their models are sometimes described as open source, but the training code has not been released nor has the training data been documented, and they do not meet the terms of either The Open Source AI Definition or the Model Openness Framework from the Linux Foundation.[citation needed]

Qwen2

Qwen2 was released in June 2024, and in September it released some of its models with open weights, while keeping its most advanced models proprietary.[10][11] Qwen2 contains both dense and sparse models.[12]

In November 2024, QwQ-32B-Preview, a model focusing on reasoning similar to OpenAI’s o1, was released under the Apache 2.0 License, although only the weights were released, not the dataset or training method.[13][14] QwQ has a 32K token context length and performs better than o1 on some benchmarks.[15] It was also in November 2024 that the Accio application was launched.[16] Accio is an AI native application that is built upon Qwen and is used to generate market insights and answer sourcing questions for Alibaba’s business to business e-commerce site. The tool is able to automate labor intensive tasks like data collection and trend tracking. [17]

The Qwen-VL series is a line of visual language models that combines a vision transformer with an LLM.[6][18] Alibaba released Qwen2-VL with variants of 2 billion and 7 billion parameters.[19][20][21]

In January 2025, Qwen2.5-VL was released with variants of 3, 7, 32, and 72 billion parameters.[22] All models except the 72B variant are licensed under the Apache 2.0 license.[23] Qwen-VL-Max is Alibaba’s flagship vision model as of 2024, and is sold by Alibaba Cloud at a cost of US$0.41 per million input tokens.[24]

Alibaba has released several other model types such as Qwen-Audio and Qwen2-Math.[25] In total, it has released more than 100 open weight models, with its models having been downloaded more than 40 million times.[11] Fine-tuned versions of Qwen have been developed by enthusiasts, such as “Liberated Qwen”, developed by San Francisco-based Abacus AI, which is a version that responds to any user request without content restrictions.[26]

On January 29, 2025, Alibaba launched Qwen2.5-Max.[27][28]

On March 24, 2025, Alibaba launched Qwen2.5-VL-32B-Instruct as a successor to the Qwen2.5-VL model. It was released under the Apache 2.0 license.[29][30]

On March 26, 2025, Qwen2.5-Omni-7B was released under the Apache 2.0 license and made available through chat.qwen.ai, as well as platforms like Hugging Face, GitHub, and ModelScope. The Qwen2.5-Omni model accepts text, images, videos, and audio as input and can generate both text and audio as output, allowing it to be used for real-time voice chatting.[31]

Qwen3

On April 28, 2025, the Qwen3 model family was released,[32] with all models licensed under the Apache 2.0 license. The Qwen3 model family includes both dense and MoE models. The sizes of the dense models include 0.6B, 1.7B, 4B, 8B, 14B, and 32B, and the MoE models include 30B-A3B (30B with 3B activated parameters) and 235B-A22B (235B with 22B activated parameters)[33]. They were trained on 36 trillion tokens in 119 languages and dialects.[34]

The Qwen3 collection include[33]:

  • Qwen3, which supports switching between thinking and non-thinking mode
  • Qwen3 Base, the pretrained base model
  • Qwen3 Instruct, which supports only non-thinking mode
  • Qwen3 Thinking, which supports only thinking mode

In additional to open-weights models, Qwen3 also includes Qwen3-Max, a proprietary large model with over 1T parameters, trained with about 36T tokens, that is available through API,[35] and Qwen3-Max-Thinking, its reasoning variant that can generate text, pictures, or video.[36]

Qwen3.5 and Qwen3.6

On February 16, 2026, Qwen3.5 and Qwen3.5-Plus were released. Qwen3.5 is open-weights.[37]

While previous models have been open source, Qwen3.5-Omni and Qwen3.6-Plus were released in April 2026 as proprietary; access to these tools is limited to the chatbots’ websites and the Alibaba cloud platform.[38]

List of models
Name Release date License Ref.
Qwen (Tongyi Qianwen) September 2023 72B, 14B, 7B: Tongyi Qianwen
1.8B: Tongyi Qianwen Research 		[39][40]
Qwen-VL August 2023 Max, Plus: Proprietary
Base, Chat: Tongyi Qianwen 		[41][42][43]
Qwen2 June 2024 72B: Tongyi Qianwen
57B-A14B, 7B, 1.5B, 0.5B: Apache 2.0 		[11][44][45]
Qwen2-Audio August 2024 Apache 2.0 [46][47]
Qwen2-VL December 2024 72B: Qwen
7B, 2B: Apache 2.0 		[19][48][49]
Qwen2.5 September 2024 72B: Qwen
32B, 14B, 7B, 1.5B, 0.5B: Apache 2.0
3B: Qwen Research 		[50][51]
Qwen2.5-Coder November 2024 32B, 14B, 7B, 1.5B, 0.5B: Apache 2.0
3B: Qwen Research 		[52][53]
QvQ December 2024 Qwen [54][55]
Qwen2.5-VL January 2025 72B: Qwen
32B, 7B: Apache 2.0
3B: Qwen Research 		[56][57][58][59][60]
QwQ-32B March 2025 Apache 2.0 [61][62]
Qwen2.5-Omni March 2025 7B: Apache 2.0
3B: Qwen Research 		[63][64][65]
Qwen3 April 2025 Apache 2.0 [66][67]
Qwen3-Coder (AKA Qwen3-Coder-480B-A35B)
Qwen3-Coder-Flash (AKA Qwen3-Coder-30B-A3B) 		July 2025 [68][69][70]
Qwen3-Max September 2025 Proprietary [71]
Qwen3-Next September 2025 Apache 2.0 [2][72]
Qwen3-Omni September 2025 [73][74]
Qwen3-VL September 2025 [75][76]
Qwen3-Coder-Next February 2026 [77][78]
Qwen3.5 February 2026 [79][80][81]
Qwen3.5-Plus February 2026 Proprietary [80]
Qwen3.6-Plus April 2026 [82]

History

In a January 2026 update of the Qwen mobile application, the Alibaba Group connected the chatbot to the company’s ecosystem, starting with food-service delivery. The company announced plans to allow users to assign ecosystem-related-tasks to more company platforms such as Taobao and Fliggy and to help consumers with errands, including phone calls and document processing[83].

Former Qwen AI model division head, Lin Jungyang, resigned in March 2026 after the release of Qwen3.5 and Qwen3.5-Plus[84], becoming the latest of three executives to have left Alibaba this year. Amid concern that this could mean a shift away from research and open-source artificial intelligence,[85] Alibaba said it will continue its focus on open source.[86]

Later the same month, a company announcement revealed the formation of a new AI business unit, Alibaba Token Hub[87]. The new unit will supervise AI-related work, led by Alibaba CEO Eddie Wu and includes Tongyi Large Model Business Unit, MaaS Business ⁠Line, Qwen, Wukong, and AI Innovation[88]. The Token Hub is further supported by senior leaders Zhou Jingren, as chief AI architect, and CTO Wu Zeming. Led by Jingren, Alibaba’s Tongyi Large Model Business Unit, formerly known as Tongyi Laboratory, was restructured to specialize in the development of Qwen AI models.[89]

Reception

There are over 200,000 variations of Qwen’s open-source AI models on the Hugging Face’s model list; Qwen3-VL-2B-Instruct has surpassed 18 million downloads globally.[90] The platform contains many small, specialized Qwen derivatives by third-party developers. Models can also be found on Alibaba’s ModelScope, which was created for greater open-source accessibility due to China’s restriction of Hugging Face in 2022.[91]

The U.S.-China Economic and Security Review Commission reported that the Chinese approach in developing open-source AI models, such as Qwen, has been critical to China’s ability to overcome constraints in compute and to expand its capabilities in real-world data curation. This enhanced capacity in data curation is perceived to be important to China’s ability to integrate AI into research, manufacturing, and robotics sectors.[91]

See also

References

  1. ^ Vidal, Nick (5 November 2025). “State of the Source at ATO 2025: State of the “Open” AI”. Open Source Initiative. Retrieved 10 April 2026.
  2. ^ a b Franzen, Carl (30 April 2025). “Qwen swings for a double with 2.5-Omni-3B model that runs on consumer PCs, laptops”. Retrieved 10 April 2026.
  3. ^ Mo, Liam; Hall, Casey (19 September 2024). “Alibaba accelerates AI push by releasing new open-source models, text-to-video”. Reuters.
  4. ^ Horwitz, Josh; Ye, Josh (11 April 2023). “Alibaba to roll out generative AI across apps”. Reuters.
  5. ^ Hall, Casey (13 September 2023). “Alibaba opens AI model Tongyi Qianwen to the public”. Reuters.
  6. ^ a b Bai, Jinze; et al. (28 September 2023). “Qwen Technical Report”. arXiv:2309.16609 [cs.CL].
  7. ^ “Qwen/techmemo-draft.md”. GitHub. August 3, 2023. Archived from the original on March 7, 2025. Retrieved March 5, 2025.
  8. ^ Fan, Feifei (2023-12-01). “Alibaba unveils new Tongyi Qianwen AI language model”. global.chinadaily.com.cn.
  9. ^ Ye, Josh (August 3, 2023). “Alibaba rolls out open-sourced AI model to take on Meta’s Llama 2”. reuters. Archived from the original on 2023-10-10. Retrieved 2024-11-29.
  10. ^ Jiang, Ben (7 June 2024). “Alibaba says new AI model Qwen2 bests Meta’s Llama 3 in tasks like maths and coding”. South China Morning Post.
  11. ^ a b c Kharpal, Arjun (19 September 2024). “China’s Alibaba launches over 100 new open-source AI models, releases text-to-video generation tool”. CNBC.
  12. ^ Yang, An; et al. (10 September 2024). “Qwen2 Technical Report”. arXiv:2407.10671 [cs.CL].
  13. ^ Dickson, Ben (29 November 2024). “Alibaba releases Qwen with Questions, an open reasoning model that beats o1-preview”. VentureBeat. Archived from the original on 18 January 2025. Retrieved 1 December 2024.
  14. ^ 故渊 (2024-11-28). “阿里通义千问 QwQ 登场：开源 AI 推理新王，MATH 测试超 OpenAI o1 模型 – IT之家”. ITHome (Chinese website).
  15. ^ Wiggers, Kyle (27 November 2024). “Alibaba releases an ‘open’ challenger to OpenAI’s o1 reasoning model”. TechCrunch.
  16. ^ “Alibaba arms merchants with AI assistant to streamline cross-border sourcing”. South China Morning Post. 2024-11-12. Archived from the original on February 11, 2026. Retrieved 2026-03-11.
  17. ^ “Alibaba’s Accio Hits 1 Million Users, Unveiling New AI Features to Empower SMEs in Global Trade”. Archived from the original on July 20, 2025.
  18. ^ Browne, Ryan (31 December 2024). “Alibaba slashes prices on large language models by up to 85% as China AI rivalry heats up”. CNBC.
  19. ^ a b Franzen, Carl (29 August 2024). “Alibaba releases new AI model Qwen2-VL that can analyze videos more than 20 minutes long”. VentureBeat. Archived from the original on 6 August 2025. Retrieved 29 April 2025.
  20. ^ 沛霖 (2024-08-30). “阿里通义千问推出 Qwen2-VL：开源 2B / 7B 参数 AI 大模型，处理任意分辨率图像无需分割成块”. ITHome (Chinese website).
  21. ^ Wang, Peng; Bai, Shuai; Tan, Sinan; Wang, Shijie; Fan, Zhihao; Bai, Jinze; Chen, Keqin; Liu, Xuejing; Wang, Jialin; Ge, Wenbin; Fan, Yang; Dang, Kai; Du, Mengfei; Ren, Xuancheng; Men, Rui; Liu, Dayiheng; Zhou, Chang; Zhou, Jingren; Lin, Junyang (September 18, 2024). “Qwen2-VL: Enhancing Vision-Language Model’s Perception of the World at Any Resolution”. Cs.CV. arXiv:2409.12191.
  22. ^ “Qwen2.5 VL! Qwen2.5 VL! Qwen2.5 VL!”. Qwen. 2025-01-26. Retrieved 2025-04-28.
  23. ^ “Qwen/Qwen2.5-VL-72B-Instruct · Hugging Face”. huggingface.co. 2025-04-28. Retrieved 2025-04-28.
  24. ^ Jiang, Ben (31 December 2024). “Alibaba Cloud cuts AI visual model price by 85% on last day of the year”. South China Morning Post.
  25. ^ Franzen, Carl (8 August 2024). “Alibaba claims no. 1 spot in AI math models with Qwen2-Math”. VentureBeat. Archived from the original on 18 January 2025. Retrieved 29 November 2024.
  26. ^ Mims, Christopher (April 19, 2024). “Here Come the Anti-Woke AIs”. WSJ. Retrieved November 29, 2024.{{cite web}}: CS1 maint: deprecated archival service (link)
  27. ^ “Qwen2.5-Max: Exploring the Intelligence of Large-scale MoE Model”. Github. 29 January 2025.
  28. ^ Baptista, Eduardo (January 29, 2025). “Alibaba releases AI model it says surpasses DeepSeek”. Reuters.
  29. ^ “Qwen2.5-VL-32B: Smarter and Lighter”. Qwen. 2025-03-24. Retrieved 2025-03-25.
  30. ^ Nikhil (2025-03-24). “Qwen Releases the Qwen2.5-VL-32B-Instruct: A 32B Parameter VLM that Surpasses Qwen2.5-VL-72B and Other Models like GPT-4o Mini”. MarkTechPost. Retrieved 2025-03-25.
  31. ^ “Qwen2.5-Omni Technical Report”. Research Gate. March 2025. Retrieved 12 March 2026.
  32. ^ Ara Shaikh, Jasmeen (April 28, 2025). “Alibaba unveils advanced Qwen 3 AI as Chinese tech rivalry intensifies”. Reuters.
  33. ^ a b “Qwen3 – a Qwen Collection”. huggingface.co. 2025-12-31. Retrieved 2026-04-07.
  34. ^ Wiggers, Kyle (28 April 2025). “Alibaba unveils Qwen3, a family of ‘hybrid’ AI reasoning models”. TechCrunch. Archived from the original on 29 April 2025. Retrieved 29 April 2025.
  35. ^ “Alibaba launches Qwen3-Max AI model with more than 1 trillion parameters”. The Hindu. 2025-09-24. ISSN 0971-751X. Retrieved 2026-03-12.
  36. ^ Cheng, Evelyn (2026-01-28). “One year after DeepSeek, Chinese AI firms from Alibaba to Moonshot race to release new models”. CNBC. Retrieved 2026-02-13.
  37. ^ Butts, Dylan (February 16, 2026). “Alibaba unveils Qwen3.5 as China’s chatbot race shifts to AI agents”. CNBC. Retrieved February 26, 2026.
  38. ^ “Chinese AI giants pivot toward proprietary models to drive revenue, performance”. South China Morning Post. 2026-04-02. Retrieved 2026-04-16.
  39. ^ Jiang, Ben (13 September 2023). “Alibaba opens Tongyi Qianwen model to public as new CEO embraces AI”. South China Morning Post.
  40. ^ “Qwen / README.md”. 20 June 2024. Retrieved 10 April 2026 – via GitHub.
  41. ^ Kharpal, Arjun (25 August 2023). “Alibaba launches AI model that can understand images and have more complex conversations”. CNBC.
  42. ^ “Introducing Qwen-VL”. Qwen. 25 January 2024. Retrieved 10 April 2026.
  43. ^ “Qwen-VL / LICENSE”. 22 August 2023. Retrieved 10 April 2026 – via GitHub.
  44. ^ “Hello Qwen2”. Qwen. 6 June 2024. Retrieved 10 April 2026.
  45. ^ “LICENSE · Qwen/Qwen2-72B”. Hugging Face. 31 May 2024. Retrieved 10 April 2026.
  46. ^ 沛霖 (2024-08-13). “阿里通义千问开源 Qwen2-Audio 7B 语音交互大模型：自由互动，无需输入文本”. ITHome (Chinese website).
  47. ^ “README.md · Qwen/Qwen2-Audio-7B”. 20 November 2024. Retrieved 10 April 2026 – via Hugging Face.
  48. ^ “Qwen2-VL: To See the World More Clearly”. Qwen. 28 August 2024. Retrieved 10 April 2026.
  49. ^ “LICENSE · Qwen/Qwen2-VL-72B”. 6 December 2024. Retrieved 10 April 2026 – via Hugging Face.
  50. ^ “Alibaba accelerates AI push by releasing new open-source models, text-to-video”. Reuters. September 19, 2024.
  51. ^ “Qwen2.5: A Party of Foundation Models!”. Qwen. 18 September 2024. Retrieved 10 April 2026.
  52. ^ Nuñez, Michael (12 November 2024). “Qwen2.5-Coder just changed the game for AI programming—and it’s free”. VentureBeat.[dead link]
  53. ^ “Qwen2.5-Coder Series: Powerful, Diverse, Practical”. Qwen. 11 November 2024. Retrieved 10 April 2026.
  54. ^ “QVQ: To See the World with Wisdom”. Qwen. 24 December 2024. Retrieved 10 April 2026.
  55. ^ “LICENSE · Qwen/QVQ-72B-Preview”. 25 December 2024. Retrieved 10 April 2026 – via Hugging Face.
  56. ^ Wiggers, Kyle (27 January 2025). “Alibaba’s Qwen team releases AI models that can control PCs and phones”. TechCrunch.
  57. ^ “LICENSE · Qwen/Qwen2.5-VL-72B-Instruct”. 27 January 2025. Retrieved 10 April 2026 – via Hugging Face.
  58. ^ “README.md · Qwen/Qwen2.5-VL-32B-Instruct”. 14 April 2025. Retrieved 10 April 2026 – via Hugging Face.
  59. ^ “README.md · Qwen/Qwen2.5-VL-7B-Instruct at c6adf136ef57b4f6a6d7897c11bd3863bda237d7”. 6 February 2025. Retrieved 10 April 2026 – via Hugging Face.
  60. ^ “LICENSE · Qwen/Qwen2.5-VL-3B-Instruct”. 27 January 2025. Retrieved 10 April 2026 – via Hugging Face.
  61. ^ Franzen, Carl (5 March 2025). “Alibaba’s new open source model QwQ-32B matches DeepSeek-R1 with way smaller compute requirements”. VentureBeat. Archived from the original on 16 May 2025. Retrieved 29 April 2025.
  62. ^ “LICENSE · Qwen/QwQ-32B”. 5 March 2025. Retrieved 10 April 2026 – via Hugging Face.
  63. ^ Butts, Dylan (27 March 2025). “Alibaba launches new open-source AI model for ‘cost-effective AI agents’“. CNBC. Retrieved 10 April 2026.
  64. ^ “LICENSE · Qwen/Qwen2.5-Omni-7B”. 26 March 2025. Retrieved 10 April 2026 – via Hugging Face.
  65. ^ “LICENSE · Qwen/Qwen2.5-Omni-3B”. 30 April 2025. Retrieved 10 April 2026 – via Hugging Face.
  66. ^ Ara Shaikh, Jasmeen (April 28, 2025). “Alibaba unveils advanced Qwen 3 AI as Chinese tech rivalry intensifies”. Reuters.
  67. ^ “Qwen3: Think Deeper, Act Faster”. 28 April 2025. Retrieved 10 April 2026.
  68. ^ “Alibaba rolls out new AI coding model Qwen3-Coder, says it’s their most powerful”. Computerworld. Retrieved 2025-07-24.
  69. ^ “LICENSE · Qwen/Qwen3-Coder-480B-A35B-Instruct”. 24 July 2025. Retrieved 10 April 2026 – via Hugging Face.
  70. ^ “LICENSE · Qwen/Qwen3-Coder-30B-A3B-Instruct”. 31 July 2025. Retrieved 10 April 2026 – via Hugging Face.
  71. ^ “Qwen3-Max: Just Scale it”. Qwen. 24 September 2025. Retrieved 10 April 2026.
  72. ^ “Qwen/Qwen3-Next-80B-A3B-Instruct · Hugging Face”. huggingface.co. 2025-09-11. Retrieved 2025-09-13.
  73. ^ Thomas, Prasanth Aby (23 September 2025). “New Alibaba model Qwen3-Omni heightens competition in multimodal AI”. Computerworld. Retrieved 11 April 2026.
  74. ^ “Qwen/Qwen3-Omni-30B-A3B-Instruct · Hugging Face”. 2025-09-22. Retrieved 2025-09-23 – via Hugging Face.
  75. ^ “Qwen3-VL: Sharper Vision, Deeper Thought, Broader Action”. qwen.ai. 2025-09-22. Retrieved 2025-11-10.
  76. ^ “README.md · Qwen/Qwen3-VL-235B-A22B-Instruct at main”. 26 November 2025. Retrieved 10 April 2026 – via Hugging Face.
  77. ^ “Qwen3-Coder-Next: Pushing Small Hybrid Models on Agentic Coding”. qwen.ai. 2026-02-02. Retrieved 2026-02-05.
  78. ^ “README.md · Qwen/Qwen3-Coder-Next”. 3 February 2026. Retrieved 10 April 2026 – via Hugging Face.
  79. ^ “Qwen3.5: Towards Native Multimodal Agents”. qwen.ai. 2026-02-16.
  80. ^ a b Butts, Dylan (February 16, 2026). “Alibaba unveils Qwen3.5 as China’s chatbot race shifts to AI agents”. CNBC. Retrieved February 26, 2026.
  81. ^ “LICENSE · Qwen/Qwen3.5-397B-A17B”. 16 February 2026. Retrieved 10 April 2026 – via Hugging Face.
  82. ^ QwenTeam. “Qwen3.6-Plus: Towards Real World Agents”. Qwen. Retrieved 4 April 2026.{{cite web}}: CS1 maint: url-status (link)
  83. ^ Qu, Tracy (2026-01-15). “China’s Alibaba Links Qwen AI App to Vast Consumer Ecosystem”. The Wall Street Journal. Retrieved 2026-04-16.
  84. ^ Pan, Che; Chen, Laurie (2026-03-04). “Alibaba forms task force to boost AI development after Qwen chief’s exit”. Reuters. Retrieved 2026-03-05.
  85. ^ Franzen, Carl (2026-03-04). “Did Alibaba just kneecap its powerful Qwen AI team? Key figures depart in wake of latest open source release”. VentureBeat. Retrieved 2026-03-05.
  86. ^ Jiaying, Li (2026-03-05). “Alibaba committed to AI advancement amid leadership shake up”. China Daily. Retrieved 2026-03-05.
  87. ^ Qu, Tracy (2026-03-17). “Alibaba Consolidates AI Operations Under New Business Group”. The Wall Street Journal. Retrieved 2026-04-16.
  88. ^ Hall, Casey (March 16, 2026). “Alibaba CEO takes helm of new AI-focused business group”. Reuters.
  89. ^ Lahiri, Anusuya. “Alibaba shakes up leadership for AI dominance as CEO Eddie Wu takes total control”. www.msn.com. Retrieved 2026-04-16.
  90. ^ “Qwen/Qwen3-VL-2B-Instruct · Hugging Face”. huggingface.co. 2025-12-31. Retrieved 2026-04-16.
  91. ^ a b Luong, Ngor (March 23, 2026). “How China’s Open AI Strategy Reinforces Its Industrial Dominance” (PDF). U.S.-China Economic and Security Review Commission.