ALIBABA OPENS QWEN3.8 WEIGHTS UNDER APACHE 2.0

The Qwen3.8-27B is a native multimodal dense model designed to deliver strong performance across multiple domains with a relatively compact parameter count. According to Alibaba, the model surpasses Qwen3.7-Plus in overall benchmarks and demonstrates particular strength in real-world coding and office workflow applications. The model supports 262K native context length, which can be extended to 1M tokens for longer document processing. By releasing the weights under Apache 2.0, Alibaba enables developers to use, modify, and distribute the models for both commercial and non-commercial purposes. The release aligns with broader industry trends of open-sourcing model weights, allowing researchers and developers to build upon the architecture without licensing restrictions. The coding performance focus addresses a key use case as AI models increasingly integrate into development workflows.