People visit the booth of Huawei at the Cairo ICT expo in Cairo, Egypt, on Nov. 17, 2024.

 

China's open-source drive is gaining fresh momentum, as a growing number of domestic tech firms step up efforts to make their self-developed technologies publicly available-a trend that analysts said is crucial for building an independent, transparent, and collaborative innovation ecosystem in the era of artificial intelligence (AI).

In the latest move, technology giant Huawei on Tuesday announced the full open-source release of its Ascend AI hardware enablement framework, known as CANN, along with its Mind series of application enablement suites and development toolkits, according to a news release posted on the company's website.

The move aims to empower developers with greater autonomy for in-depth optimization and customized development, accelerating innovation across the developer community and making the Ascend ecosystem more robust and user-friendly, noted the release.

Just a day earlier, smartphone maker Xiaomi officially open-sourced its MiDashengLM-7B, a self-developed audio large language model that achieved state-of-the-art results across 22 public benchmark datasets, the company said in a statement on its Weibo account.

The back-to-back announcements add fresh impetus to China's open-source drive, which will be crucial for building a resilient and self-sustaining AI innovation ecosystem amid intensifying global competition, and advancing shared progress in technology worldwide, veteran telecom industry analyst Ma Jihua told the Global Times on Tuesday.

Huawei's announcement was made during the Ascend Computing Industry Summit held in Beijing on Tuesday, where Rotating Chairman Eric Xu said that computing power remains at the core of Huawei's AI strategy. 

At the summit, Huawei and a group of industry partners, academic institutions, and research organizations jointly launched a new initiative to co-develop the CANN open-source ecosystem, in a bid to pool industry strength to build an innovative Ascend ecosystem, per the release.

Standing for Compute Architecture for Neural Networks, CANN is a heterogeneous computing architecture developed by Huawei for AI scenarios. It is designed to bridge the gap between upper-layer AI frameworks and lower-level processors and programming environments, serving as a key platform to enhance the computing efficiency of Ascend AI processors.

By open-sourcing it in full, Huawei is offering developers deeper access to optimize AI workloads and build custom applications, further enhancing the Ascend ecosystem's adaptability and scalability, according to Ma.

In recent years, Huawei has open-sourced several of its core self-developed software platforms, including the OpenHarmony operating system, openEuler, OpenGauss database, MindSpore AI framework, and CarbonData for big data and middleware. 

Xiaomi's MiDashengLM-7B is another notable addition to China's rapidly expanding portfolio of open-source large language models. The company said in Monday's statement that it will be widely applied across scenarios such as smart cockpits and smart homes, enabling more precise and seamless interaction. 

Other major players, including Alibaba, are ramping up their efforts. In late July, the e-commerce and cloud giant unveiled Qwen3-Coder, an open-source AI coding model under its Tongyi Qianwen family, demonstrating strong performance in code generation. This model ranks among the top open-source models globally, according to Science and Technology Daily.

The surge in open-source initiatives reflects a broader push by China to bolster the country's software infrastructure and enhance self-reliance in technological innovation.

In a recent meeting with industry participants, officials from the Ministry of Industry and Information Technology stressed the need to enhance China's open-source infrastructure and accelerate the development of a national-level AI open-source community.

The ministry previously highlighted the steady advancements of China's open-source ecosystem, with a wave of new projects launched by technology companies, including the Cangjie programming language, Galaxy auto operating system, and Wenxin large language models.

As China's open-source community matures, the interplay among tech giants, start-ups, academia, and regulators will be key to shaping the next generation of foundational software and AI models that underpin the country's long-term digital transformation, Ma said.