Qwen3.8: three language models, three licenses, and a fourth for images
Alibaba's Qwen3.8 models ship under Apache 2.0, the Qwen3.8-Max License and Qwen Community 1.0, while Qwen-Image-2.1 is research-only. A map of what each license lets you do.
Alibaba’s Qwen team published three open-weights language models in August 2026, and each came with different terms. Many people still treat “Qwen” as shorthand for “Apache 2.0”, which is no longer true for the whole family. This article is a reading aid, not legal advice.
The family at a glance
| Model | Weights released | Size | Context | Input | License |
|---|---|---|---|---|---|
| Qwen3.8-Max | 12 Aug 2026 | 2.4T total, 95B active (mixture of experts) | 1M tokens | Text, image, video | Qwen3.8-Max License |
| Qwen3.8-27B | 14 Aug 2026 | 27.8B (dense) | 262K native, extendable to 1M | Text, image, video | Apache 2.0 |
| Qwen3.8-Flash-Next | 26 Aug 2026 | 180B total, 6B active (mixture of experts) | 262K native, extendable to 1M | Text, image | Qwen Community License 1.0 |
The flagship was announced on 2 August with a hosted API, and its weights followed ten days later. Flash-Next is an early preview of the architecture Alibaba plans to use for Qwen4.
What each license asks for
| License | Credit line | Hosting and assistant businesses | Everyone else |
|---|---|---|---|
| Apache 2.0 | Keep the license and notices | No special conditions | Free to use commercially |
| Qwen3.8-Max License | Show the model name in the interface of a product with more than 100 million monthly users or US$20 million of monthly revenue | Separate license from Qwen if combined revenue exceeds US$50 million in any 12 months | Free to use, modify, host, fine-tune and sell |
| Qwen Community License 1.0 | Same credit rule as Max | Separate license from Qwen first, with no revenue threshold | Free to use, modify, host, fine-tune and sell |
| Qwen Research License | Keep the attribution notice | Not allowed without a separate license | Research or evaluation only |
“Hosting and assistant businesses” means a model-as-a-service business that sells API or fine-tuning access, or an “AI work assistant” business, which Qwen defines as an independent coding or office-productivity product. Using the model inside your own product, or relaying requests to a model hosted by someone else, is not covered by the clause. Our analysis of hosting clauses compares these rules with those of other developers.
Three things to take away
- The smallest model is the most permissive. Qwen3.8-27B is the only one of the three with no credit or hosting conditions at all. It also fits on a single GPU once compressed, so for most people it is the practical choice.
- The preview model has the strictest hosting rule. Qwen3.8-Flash-Next applies the hosting condition at any revenue level, while the much larger Qwen3.8-Max applies it only above US$50 million.
- The name tells you nothing about the license. Check each model’s LICENSE file. Qwen-Image-2.1, released on 20 September, is research-only, while earlier Qwen image models were Apache 2.0. Qwen3-Embedding and Qwen3-TTS are Apache 2.0.
For a hands-on look at the permissive option, see our comparison of Gemma 4 and Qwen3.8-27B. AI licenses explained walks through what the terms in these tables mean.
Sources
- Qwen3.8 repository and release news (GitHub) github.com
- Alibaba Unveils Qwen3.8-27B and Releases Weights of Qwen3.8 Flagship Model (Alibaba Cloud, 17 Aug 2026) alibabacloud.com
- Qwen3.8-Max License (Hugging Face) huggingface.co
- Qwen Community License 1.0 (Qwen3.8-Flash-Next) huggingface.co
- Qwen3.8-27B model card (Hugging Face) huggingface.co
- Qwen-Image-2.1 license (GitHub) github.com
Models in this story
| Model | Released | Parameters | Context | License | Smallest variant needs |
|---|---|---|---|---|---|
| Qwen3.8-Max Alibaba Qwen | 2.45T 95B active | 1M | Qwen3.8-Max License (Conditional) | 1.69 TB Multi-node cluster | |
| Qwen3.8-27B Alibaba Qwen | 27.8B | 256K | Apache 2.0 (Permissive) | 19 GB Gaming GPU | |
| Qwen3.8-Flash-Next Alibaba Qwen | 180B 6B active | 256K | Qwen Community 1.0 (Conditional) | 124 GB Big-memory desktop | |
| Qwen-Image-2.1 Alibaba Qwen | 7.1B | – | Qwen Research License (Non-commercial) | 20 GB Gaming GPU |