Huawei Opensource Pangu 92-Billion-Parameter Model openPangu-2.0-Flash
Huawei has officially open-sourced the 92-billion-parameter openPangu-2.0-Flash model. Based on Ascend native training and inference technology, it aims to build an intelligent foundation for the Agent era. Model weights, basic inference code, and training/inference operators are now available on the open-source platform. The Pro version will be released in July, with more components to follow in the second half of the year.
IT Home, June 30 news: Huawei officially announced that the Huawei open-source Pangu 92-billion-parameter openPangu-2.0-Flash model was officially launched on the open-source platform today (the 30th).
According to the introduction, open-source Pangu openPangu is Huawei's open-source AI model brand. It is committed to providing best practice references for the industry to utilize Ascend effectively through Ascend native training and inference technology, helping to build a powerful intelligent foundation for the Agent era, and accelerating AI business innovation and ecosystem prosperity. Components related to the openPangu-2.0 model will be open-sourced starting June 30: - openPangu-2.0-Flash model weights, basic inference code, and training/inference operators will be officially launched on the open-source platform on June 30. - openPangu-2.0-Pro model weights and basic inference code will be launched on the open-source platform in July. - More open-source components will be launched successively in the second half of the year.
IT Home attaches the relevant open-source information link: https://gitcode.com/ascend-tribe
During the Huawei Developer Conference HDC 2026 earlier this month, Yu Chengdong, Huawei Executive Director, Director of the Product Investment Review Committee, and Chairman of Terminal BG, officially released the open-source Pangu openPangu 2.0.
The openPangu 2.0 model features a 512K context window and includes two versions: Pro and Flash: - 2.0 Pro: Total parameters 505B, Active parameters 18B - 2.0 Flash: Total parameters 92B, Active parameters 6B
openPangu 2.0 plans to open-source 7 major components starting June 30, including newly open-sourced pre-training code, post-training code, and training operators.
Related reading:
Advertising Statement: External jump links contained in the text (including but not limited to hyperlinks, QR codes, passwords, etc.) are used to convey more information and save selection time. Results are for reference only. All IT Home articles contain this statement.
This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.