XingChen-AGI/Xing4.0-29B-A4B
Xing4.0-29B-A4B releases open weights for a 29B parameter MoE model with 4B active parameters, supporting 256K context. Developed by China Telecom, it targets agent workflows and complex reasoning. The model is compatible with vLLM, SGLang, and Transformers, and was trained on Ascend NPUs. Developers can download the weights and integrate the model into existing inference stacks for specialized domain tasks.
README
XingChen-AGI/Xing4.0-29B-A4B View on Hugging Face
Loading the README from Hugging Face…