88Z AI, 30 June — Huawei has announced the open-source release of its 92-billion parameter openPangu-2.0-Flash model. Built on Ascend native technology, it aims to provide foundational support for agent applications. Model weights, inference code, and key operators are now available on open-source platforms, with the Pro version expected in July and additional components rolling out in the second half of the year.
At a glance
- Huawei open-sources the 92B parameter openPangu-2.0-Flash model based on Ascend native technology.
- Model weights and code are available now, with the Pro version launching in July and more components in the second half of the year.
The update
Huawei has officially open-sourced the openPangu-2.0-Flash model, which features 92 billion parameters. Leveraging Ascend native training and inference technologies, this release aims to build a solid intelligent infrastructure foundation for the emerging era of agents. Developers can currently access the model weights, basic inference code, and related training and inference operators on open-source platforms. According to the roadmap, the more capable Pro version is scheduled for release in July. Furthermore, Huawei plans to roll out additional components in the second half of the year to further enhance the ecosystem and meet diverse development needs.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
- 华为Source published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.