Qwen Announces Upcoming Release of Open-Weights Model "3.8" Next Week
Alibaba Cloud's Qwen team has teased the upcoming release of a new open-weights model, referred to as "3.8", scheduled for next week. This announcement has generated significant anticipation within the local LLM community on platforms like Reddit. Qwen models are highly regarded in the open-source AI ecosystem for their strong performance, and a new open-weights release allows developers to run, study, and customize the model locally. This release could provide a highly capable, smaller-sized model option for resource-constrained environments. While the exact specifications of the "3.8" model (such as whether it refers to a 3.8 billion parameter size or a version number) remain unconfirmed, the release is expected to follow Qwen's open-weights distribution model. Users will be able to download the model parameters directly from platforms like Hugging Face.
## BACKGROUND
Qwen is a family of large language and multimodal models developed by Alibaba Cloud. An open-weight model differs from fully open-source AI as it provides the final trained parameters (weights) for local execution, but may not share the complete training dataset or codebase.