Language model
Qwen 3.8 Flash Next
Alibaba
- Released
- August 26, 2026
- Data date
- September 26, 2026
Qwen 3.8 Flash Next is an open experimental architecture checkpoint, not simply a smaller Qwen3.8 model. Qwen replaces Gated Attention with Qwen Sparse Attention and combines it with Gated DeltaNet. The model has 125 billion parameters, activates 6 billion, and adds 51 billion n-gram embeddings. It accepts text, images, and video and uses Thinking mode by default.
Native context reaches 262,144 tokens and can be extended to 1 million. Thinking can be disabled through the API. The official serving model Qwen 3.8 Flash is based on this Next checkpoint according to its model card and adds 1M context plus built-in tools by default.
Specifications and access
| Specification | Value and source |
|---|---|
| Weights | Open weightsSource |
| Access | APISource |
| Parameters | 125 billionSource |
| Active parameters | 6 billionSource |
| Context window | 262,144 tokensSource |
| Maximum output | 131,072 tokensSource |
| License | Qwen Community License 1.0Source |
| Inputs | Text, Image, VideoSource |
| Outputs | TextSource |
| Reasoning | YesSource |
| Tool calling | YesSource |
Popularity
59.6 / 100
Sources and data date
| Type | Evidence and data date |
|---|---|
| Additional source | models.dev catalog (retrieved September 26, 2026) |
| Additional source | Hugging Face · Qwen3.8-Flash-Next model card · Editorial description reviewed October 2, 2026 |
| Additional source | Qwen3.8 Flash Next announcement · Editorial description reviewed October 2, 2026 |
| Additional source | OpenRouter: Popularity index (derived), CC BY 4.0 |
| Additional source | Vercel AI Gateway: Popularity index (derived), CC BY 4.0 |
| Additional source | Creative Commons Attribution 4.0 |