Alibaba · Open-weight
Checked as of 2026-10-23
Qwen
The widest size range of any open family, from 27B up to a 2.4-trillion-parameter model.
At a glance
- Licence
- Open, restricted Apache-2.0 / Qwen Community License 1.0 / Qwen3.8-Max License
Varies by model. Qwen3.8-27B and Qwen3.6 are Apache-2.0. Qwen3.8 Flash-Next and 2.4T need a separate licence for model-as-a-service or AI-assistant businesses, and large products must display the model name.
- Access
- APIHF
- Hugging Face
- huggingface.co/Qwen
- Source
- https://huggingface.co/Qwen
Variants 4
| Model | Params | Context | Hardware tier | Gated | Inputs | Status |
|---|---|---|---|---|---|---|
Qwen3.8 2.4T-A95B
Qwen/Qwen3.8-2.4T-A95B
Qwen3.8-Max License. Extensible to 1,010,000 tokens. |
2400B95B active | 262.144K | Multi-GPU ml.p5.48xlarge and up | No | text | current |
Qwen3.8 Flash-Next
Qwen/Qwen3.8-Flash-Next
Qwen Community License 1.0. 180B includes 51B of n-gram embeddings. |
180B6B active | 262.144K | Multi-GPU ml.p5.48xlarge and up | No | text, image, video | current |
Qwen3.8 27B
Qwen/Qwen3.8-27B
Dense, Apache-2.0. Safetensors total is 27.8B. |
27B | 262.144K | 80GB GPU ml.p4de.24xlarge / ml.p5.48xlarge | No | text, image, video | current |
Qwen3.6 35B-A3B
Qwen/Qwen3.6-35B-A3B
April 2026, Apache-2.0. Still heavily downloaded. |
35B3B active | 262.144K | 80GB GPU ml.p4de.24xlarge / ml.p5.48xlarge | No | text, image, video | legacy |
"—" means the figure could not be verified from a primary source on 2026-10-23 and has been left blank rather than guessed. Hardware tier is from total parameters (all experts for mixture-of-experts models), since every weight must sit in memory. SageMaker rates on the open-weight page.
Strengths
- Sizes from small dense models to a 2.4T open-weight mixture-of-experts model.
- 262,144-token native context, which can be extended to about 1M with YaRN.
- Most mid-size models take text, image and video natively.
- No Hugging Face gating on any listed repo.
Watch out for
- Licences differ within one generation. Read the licence for the exact model, not the family.
- Bedrock and JumpStart carry only older Qwen3 models, none of the listed ones.
- The 2.4T model is text-only and needs several 8-GPU nodes.