Xing4.0-29B-A4B: Chinese Model Optimized for Ascend NPU

XingChen-AGI published Xing4.0-29B-A4B on Hugging Face: a 29B-parameter model with 4B activated per token, native 256K (expandable to 512K) context length, and trained on Ascend NPUs using MindSpore with claimed efficiency gains.

• Swift Qwen 3.8 27B Surpasses 100k Downloads2026-09-17
• NVIDIA advances native GPU programming in Rust2026-09-17
• Censorship Begins on HuggingFace2026-09-15
• Aurora1.0-150M Released2026-09-14
• Deepseek V4.1 Flash is 748B total/552B base2026-09-10