AI BriefingMS-Swift Adds Support for Qwen 3.8 MoE
Key point
The latest commit to the MS-Swift GitHub repository confirms support for the Qwen 3.8 35B MoE model.
Details
Support for the Qwen 3.8 model family has been added via the latest commit (ab726e9) to the modelscope/ms-swift GitHub repository.
The confirmed model utilizes a MoE (Mixture-of-Experts) architecture with a total of 35B parameters and approximately 3B active parameters. This is the same structure as the Qwen3.6-35B-A3B, which was previously noted for its efficiency in 8–16GB VRAM GPU environments.
With this update, developers can now utilize the latest Qwen MoE models within the ms-swift framework.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Apple Releases LensVLM-9B for Processing Compressed Text ImagesJevper Released: A Jev Classification Wrapper for OpenAI-Compatible Models with Automatic Logprobs Detection