Skip to content
Live
Hugging Face Daily Papers·· 21 hours ago

TeleChat to Xing4: Advancing Frontier Agentic Models on Domestic Chips and Frameworks

TeleChat to Xing4: Advancing Frontier Agentic Models on Domestic Chips and Frameworks

Brief overview

Xing4 is a series of LLMs developed on Huawei Ascend chips and the MindSpore framework, expanding the capabilities of the TeleChat series through the synergistic progress of architecture and training infrastructure, supporting the pre-training of 285B parameter models on 8192 blocks of Ascend 910C GPUs, and introducing the TeleGym unified agent execution platform to optimize strategies.

Article

Published on Nov 17

#1 Paper of the day

Authors:

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

Abstract

We present Xing 4, a series of large language models that extends the TeleChat lineage through coordinated advances in model architecture and training infrastructure on Ascend. This report describes the design principles and systems engineering behind the series, spanning architecture selection, large-scale pretraining, and reasoning and agentic post-training. We study how sparse activation, attention design, model depth and width, and speculative decoding balance model capacity with computation, memory access, and execution overhead. Our pretraining infrastructure builds on MindSpore HyperParallel, combining topology-aware hybrid sharding, efficient distributed optimizer execution, and coordinated management of long-sequence workloads and activation memory. CPU affinity and core isolation further reduce host scheduling interference. Together, these mechanisms support the pretraining of a 285B-parameter model on 8,192 Ascend 910C cards, providing a concrete setting for the architectural and systems studies. For post-training, we introduce TeleGym, a unified agent execution platform that connects execution, evaluation, and trajectory construction to policy optimization. Disaggregated asynchronous execution, bounded oversampling, and partial rollout accommodate heterogeneous task durations while preserving trajectory continuity across policy updates. Together, these advances establish an Ascend-oriented training foundation for the continued development of Xing 4 across model scales and increasingly demanding reasoning and agent workloads.

View arXiv page View PDF Add to collection

Get this paper in your agent:

hf papers read 2611.22115

Don't have the latest CLI?

curl -LsSf https://hf.co/cli/install.sh | bash

Models citing this paper 0

No model linking this paper

Cite arxiv.org/abs/2611.22115 in a model README.md to link it from this page.

Datasets citing this paper 0

No dataset linking this paper

Cite arxiv.org/abs/2611.22115 in a dataset README.md to link it from this page.

Spaces citing this paper 0

No Space linking this paper

Cite arxiv.org/abs/2611.22115 in a Space README.md to link it from this page.

Collections including this paper 0

No Collection including this paper

Add this paper to a collection to link it from this page.

Source: Hugging Face Daily Papers · huggingface.co