Qwen 3.8: Alibaba Releases Open Weights for AI Models

Qwen 3.8 open weights are now available, marking another significant step for Alibaba’s AI division. The Qwen team announced the availability of open weights for the Qwen 3.8 model family, including the Qwen3.8-27B and a larger version, the Qwen3.8-2.4T-A95B (Max-level) model. This follows the introduction of Qwen 3.8 Max, Alibaba's largest AI model to date. The company says Qwen3.8-Max uses 2.4 trillion parameters, with 95 billion active during processing through a Mixture of Experts architecture. Alibaba claims its performance rivals leading AI models from companies like OpenAI and Anthropic, especially in coding and complex enterprise workflows. What's interesting is Qwen3.8-Max's multimodal capability, allowing it to understand and work with text, image, and video inputs. This model is designed for various enterprise and developer uses, such as programming, office productivity, and document analysis. Alibaba is focusing on real-life applications over just high benchmark scores. The open-weight strategy for Qwen 3.8 is key. Developers can now access the open weights for Qwen3.8-27B, a multimodal model with 27 billion parameters. Alibaba states this model outperforms Qwen3.7-Plus in overall performance, particularly in coding and office tasks. The bottom line is that these new open weights provide accessible, high-performance AI tools for developers and businesses globally.