Huawei Open-Sources Pretraining, SFT and Post-Training RL Code for openPangu-2.0
Bubble signal: 0 · NeutralRelevance 32ModelsChina
Source: IT Home · Published 2026-09-28
Summary
On September 28, Huawei announced it has open-sourced the pretraining, SFT and post-training RL code for its openPangu-2.0 model, with related components rolling out on open-source platforms. openPangu is Huawei's open-source AI model brand, focused on Ascend-native training and inference as reference practice for the industry.
Bubble analysis
This is a model-code release, i.e. technical ecosystem building; it carries no investment figures, valuations or compute purchases, so it offers almost no direct evidence on whether AI is in an investment bubble. At most it indirectly shows Huawei pushing its Ascend ecosystem to reduce reliance on Nvidia, but there is no money figure here to support any judgment.
#huawei#ascend#open-source#llm
Read original ↗Related signals
linked by shared tags
- Ant Ling Releases Ling-3.1-flash Model: ~560B Total Parameters, 25B Activated Parameters#open-source#llm
- DeepSeek Open-Sources Huawei Ascend Compute Infrastructure Components, Matching Nvidia's Platform#huawei#ascend
- DeepSeek Open-Sources Infrastructure Components for Huawei Ascend Compute Platform, Matching Its Nvidia Offerings One-for-One#huawei#ascend
- Breaking the Three Walls of Compute, Storage and Communication: Huawei's Ascend Super Node Offers the 'Optimal Solution' for Training and Inference of 10-Trillion-Parameter Models#huawei#ascend
- Why Is DeepSeek Embracing Huawei Now?#huawei