Physix Frontier · News Briefing Card (IT Home · Sep 15, 2026)

China Releases Corpus 4.0, a 120GB Chinese Internet Dataset

KEY FACTS

The dataset was officially released on September 15 at a sub-forum of the Jinan Cybersecurity Awareness Week.
With a total volume of 120GB, it is now available on the Chinese Internet Corpus Resource Platform.
Guided by the Cyberspace Administration of China (CAC), the project was co-built by an association and six entities, including Baidu and iFlytek.
Users must complete a registration and verification process to download or access the data.

KEY DATA

120GBData Volume

PHYSIX OBSERVATION

The official-led iteration of high-quality Chinese corpora to version 4.0 signals that domestic large model training is shifting from 'wild growth' to a phase of 'compliance and trustworthiness.' Deep involvement from industry leaders addresses pain points in data copyright and quality while establishing sector barriers. For smaller developers, although access thresholds exist, the pathway is clear; this move promises to reduce reliance on overseas data and advance the autonomous controllability of the domestic AI ecosystem.

Source: IT Home report