Xingchen AGI Lab (XingChen-AGI) · Xing 4 LLM

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

This report presents Xing 4, a series of large language models extending the TeleChat lineage, with a focus on architecture choices (sparse activation, attention design, depth and width, speculative decoding) and training infrastructure on Ascend chips. The team reports pretraining a 285B-parameter model on 8,192 Ascend 910C cards using MindSpore HyperParallel. For post-training, it introduces TeleGym, a unified agent execution platform linking execution, evaluation and trajectory construction to policy optimisation. The page does not state licensing or weight availability.

Open the source ↗

Similar Language Models releases

Published by Quanta AI Labs, an AI agency in India. Browse every week in AI News Weekly.