OpenModelsHQ

Alibaba Qwen

Hangzhou, China · Alibaba Group · China Desk

Alibaba Qwen is based in Hangzhou, China and is part of Alibaba Group. We track 19 open-weight models from it, and the newest is Qwen3.8-Flash-Next, released 24 Aug 2026.

Models tracked
19
LLM 7 · Vision 10 · Speech 1 · Embedding 1
Newest release
Qwen3.8-Flash-Next
24 Aug 2026
Country
China
Releases timeline
19 Alibaba Qwen releases from 16 Sep 2024 to 24 Aug 2026. The largest is Qwen3.8-Flash-Next at 180B parameters. Dot area follows total parameters; a ring means the size is not published.
Releases timeline: data
ModelCategoryReleasedParameters
Qwen3.8-Flash-NextVision24 Aug 2026180B
Qwen3.8-27BVision5 Aug 202627.8B
Qwen3-ASR-0.6B-hfSpeech26 Jun 2026782M
Qwen3.6-27BVision21 Apr 202627.8B
Qwen3.6-35B-A3BVision15 Apr 202636B
Qwen3.5-2BVision28 Feb 20262.3B
Qwen3.5-4BVision27 Feb 20264.7B
Qwen3.5-9BVision27 Feb 20269.7B
Qwen3-VL-8B-InstructVision11 Oct 20258.8B
Qwen3-VL-4B-InstructVision11 Oct 20254.4B
Qwen3-Embedding-0.6BEmbedding3 Jun 2025596M
Qwen3-32BLLM27 Apr 202532.8B
Qwen3-14BLLM27 Apr 202514.8B
Qwen3-8BLLM27 Apr 20258.2B
Qwen3-0.6BLLM27 Apr 2025752M
Qwen2.5-VL-7B-InstructVision26 Jan 20258.3B
Qwen2.5-1.5B-InstructLLM17 Sep 20241.5B
Qwen2.5-7B-InstructLLM16 Sep 20247.6B
Qwen2.5-0.5B-InstructLLM16 Sep 2024494M

Origin

Alibaba launched Qwen as a beta in April 2023 under the name Tongyi Qianwen, and opened it to the public in September 2023 after regulatory clearance. The models are built by Alibaba’s Qwen team, part of Alibaba Group.

How it got here

The first open weights came in 2023: Qwen-7B on 3 August, Qwen-14B on 25 September and Qwen-72B on 30 November. Qwen2 came in June 2024. Qwen2.5 arrived on 19 September 2024 and added 3B, 14B and 32B sizes to the line-up.

Qwen3, released on 29 April 2025, had two mixture-of-experts models, Qwen3-235B-A22B and Qwen3-30B-A3B, plus six dense models from 0.6B to 32B parameters. The Qwen3.5 and Qwen3.6 generations came next. In August 2026 the team released Qwen3.8, and the model card for Qwen3.8-27B calls it the most capable generation in the Qwen open-model family to date.

Licensing pattern

The earliest Qwen weights used the Tongyi Qianwen license, and commercial deployment needed approval through an application form. The source code itself is under Apache 2.0.

The weights then moved towards Apache 2.0 as well. The Qwen2.5 repository says its open-weight models are licensed under Apache 2.0, though Wikipedia describes the Qwen2.5 family as having mixed licensing, so the license on each model card is the one to trust. The Qwen3 dense models were released under Apache 2.0, and the Qwen3.8-27B card lists it too.

One release breaks the pattern. Wikipedia reports that Alibaba previewed a 2.4 trillion parameter model, Qwen3.8-Max, in July 2026, and released its weights as Qwen3.8-2.4T-A95B on 12 August 2026. Its license requires model providers that generate more than US$50 million in revenue within 12 months to obtain a commercial license from Alibaba.

Release lineage

Newest first
  1. Qwen3.8-Flash-Next180B
  2. Qwen3.8-27B27.8B
  3. Qwen3-ASR-0.6B-hf782M
  4. Qwen3.6-27B27.8B
  5. Qwen3.6-35B-A3B36B
  6. Qwen3.5-2B2.3B
  7. Qwen3.5-4B4.7B
  8. Qwen3.5-9B9.7B
  9. Qwen3-VL-8B-Instruct8.8B
  10. Qwen3-VL-4B-Instruct4.4B
  11. Qwen3-Embedding-0.6B596M
  12. Qwen3-32B32.8B

Largest first. The last line on each tile is the memory of the smallest device that fits the model at 4-bit.