Qwen3.8-27B
Qwen3.8-27B is a dense 27-billion-parameter vision-language model released under Apache 2.0. Its card lists hosted access as coming soon; the weights are available now.
Technical details
Thinking is enabled by default and can be disabled. The official collection also provides an FP8 checkpoint.Qwen3.8-27B model card
Practical use
Practical guidance. Use this release when you need published weights under Apache 2.0 and a dense model is suitable for your hardware. Confirm that your inference runtime supports its attention implementation before comparing throughput.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | Qwen/Qwen3.8-27BQwen/Qwen3.8-27B-FP8 |
| Architecture | 64 layers; gated delta networks alternate with gated attention.Qwen3.8-27B model card |
| Context | 262,144 native tokens, extendable to one million.Qwen3.8-27B model card |
Related models
- Qwen3.8 Max
Alibaba · Qwen3.8
Hosted multimodal model with text, image and video input and a one-million-token context.
Launch date not confirmedAPI - Qwen3.8-2.4T-A95B
Alibaba · Qwen3.8
Text-only mixture-of-experts model with 2.4 trillion total parameters and 95 billion active parameters.
Launch date not confirmedDownloadable