AI Briefingllama.cpp adds support for Xing4.0-29B-A4B
·2026.11.06 03:30
Key point
The 29B parameter model features 4B activated parameters per token and supports a native 256K context length.
Details
- llama.cpp Pull Request #31589 adds support for Xing4.0-29B-A4B, a model developed by China Telecom Artificial Intelligence Technology Co., Ltd.
- Xing4.0-29B-A4B has 29B total parameters with 4B activated per token and natively supports a 256K context length, extensible to 512K.
- The model was trained entirely on the Ascend NPU platform using the MindSpore framework, achieving approximately 96% training throughput improvement over out-of-the-box performance.
- It utilizes an mHC + MLA + MTP architecture and supports inference via SGLang, vLLM, and KTransformers, as well as fine-tuning via LLaMA-Factory and MindFormers.
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
Herdr releases native desktop app for managing terminals and coding agents
Rivet Launches Open-Source Actor Platform for Scaling AI Agents on S3