The model can run locally on a laptop and demonstrates how Alibaba plans to continue pioneering open source.
August 17, 2026
In another sign that open model competition continues to heat up between Chinese and U.S. model makers, Alibaba Cloud introduced its latest variant of its flagship model, Qwen 3.8-27B.
Released over the weekend under the Apache 2.0 license, the open-weight model is optimized for software engineering, reasoning and performing long-horizon tasks. Alibaba said the model can provide stronger autonomous planning and enable flexible control of reasoning, enabling users to control how deeply they want the model to reason and whether the context used to help the model reason is kept or deleted.
Qwen 3.8 27B comes a few days after U.S. vendors Meta and Nvidia introduced their own new open models. The release intensifies the open model race between the U.S. and China while also reflecting the sustained push of AI to edge devices. Like Meta Muse Glimmer and Nvidia’s Nemotron 3.5 Lightning, the smaller 27B variant of Qwen -- whose most powerful model, Qwen 3.8-Max, with 2.4 trillion parameters, is 88 times bigger than Qwen 3.8 -- is designed to run locally on a laptop.
