Skip to content
AI IntelligenceAug 4, 2026AI Intelligence
Article

Huawei open-sourced openPangu-2.0-Pro, a 505B-parameter model with 180B sparse activation and a 512K context window.

It is the first frontier-scale model trained entirely on Ascend NPUs, and the release includes weights, inference code, and a technical report.

Data Cube AI EditorialSource: Pandaily
01

Source Brief

Huawei open-sourced openPangu-2.0-Pro, a 505B-parameter model with 180B sparse activation and a 512K context window. It is the first frontier-scale model trained entirely on Ascend NPUs, and the release includes weights, inference code, and a technical report.