Qwen/Qwen3.8-27B is an open-source multimodal model hosted on Hugging Face by Qwen and released under the Apache 2.0 license. It belongs to the Qwen3.8 family, described as the most capable generation in the Qwen open-model line. The model is a dense, deployment-friendly 27B parameter vision-language system that can understand both images and videos, with flexible thinking control for multi-step tasks. According to the model card, Qwen3.8 builds on the Qwen3.5 architecture and delivers major improvements in coding, professional work, research, and long-horizon agentic tasks.
The page emphasizes that the repository contains model weights and configuration files in Hugging Face Transformers format, and that the model is compatible with tools such as Transformers, vLLM, SGLang, TokenSpeed, and other related inference frameworks. It also notes that a hosted Qwen Cloud version is planned, offering production features such as a default 1M context length and built-in tools.
The page provides practical usage examples for several deployment options. These include loading the model with Transformers using either a pipeline or direct model/processor calls, serving it with vLLM or SGLang, and running it through Docker or Docker Model Runner. It also points users to quantized versions for use in apps like llama.cpp, Ollama, and LM Studio.
Overall, the page serves as a model hub entry, combining a brief product overview, deployment instructions, and links to benchmarks, best practices, and citation information.
Article timeline
Build this article’s timeline
Extract dated events, key actors, and open questions using the same timeline analysis available in Signal.
Comments
Login to view and post comments.