Qwen3.8-27B
Description
Qwen3.8-27B is an open language model from Qwen, released on 2026-08-05. It is the compact, "deployment-friendly" model of the Qwen3.8 generation, which the authors call the most capable generation of their open models so far. Unlike the giant models of this generation, it has 27 billion parameters and all of them work on every step, so it is much easier to run than the largest open models.
According to the model card, it understands text, images and video, including hour-long videos. It is built for programming, professional work, research and long multi-step tasks done by an AI agent. Thinking mode is on by default and can be switched off for a single request; the depth of reasoning can be adjusted, and the reasoning from earlier messages can be kept in a long conversation. The context window is 262,144 tokens and can be extended to one million.
The weights are public without any acceptance step, and the license is Apache-2.0, which allows commercial use. We found no territorial restrictions in the license. The model works with Transformers, vLLM, SGLang and other tools, and a version in the lighter FP8 format is published next to it. The card does not state how much video memory is needed. It also notes that a hosted version from the authors is planned.
Date of check: 2026-10-21.
Pros and cons
Pros
- Apache-2.0 license, weights open without an acceptance step
- Compact: 27 billion parameters, all working on each step
- Understands text, images and video
- Thinking mode that can be switched off, with adjustable depth
- Context of 262,144 tokens, extendable to one million
Cons
- The card does not state hardware requirements
- No audio input
- Weaker than the largest open models of the same generation
- Long contexts above the native size need extra settings
- Benchmark figures come from the authors, not from independent tests
Key facts
- License
- Apache-2.0
- Commercial use
- Allowed
- Developer
- Qwen
- Interface languages
- English
- Checked on
- Oct 21, 2026