Zhongzhi FlagOS completes full-factor AI training validation

MarsBitNews · 2026-03-26T06:17:15+00:00

The ZhiYuan Research Institute, in collaboration with partners, has launched the open-source software stack ZhiZhi FlagOS. It has successfully completed end-to-end training validation on six AI chips and three AI large models, allowing users to achieve high-quality training experiences across diverse hardware and reducing dependence on a single hardware platform.

MarsBitNews

2026-03-26 06:17:15

Abstract generation in progress

Mars Finance News, March 26 — The open-source unified software stack, Zhiyuan Research Institute’s joint ecosystem partner project called FlagOS, has completed end-to-end training validation on six AI chip manufacturers, three types of AI large models (language, multimodal, embodied), and five homogeneous and heterogeneous KCal clusters.

In terms of hardware compatibility, FlagOS successfully completed end-to-end training validation on six mainstream AI chips: Tianzhi Zhixin, Muxi, Cambrian, Hygon, Mooresoft, and Kunlun Chip. During language model training, the training results on each chip are highly consistent with international mainstream platforms, allowing users to achieve the same high-quality training experience across diverse hardware, significantly reducing dependence on a single hardware system. (Wide-angle observation)

View Original

This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.