On March 15, DeepZang, a Tibetan large language model with its intelligent applications, was launched in Lhasa, Xizang. This is the first Tibetan large language model in China to have completed national registration for generative artificial intelligence (AI).

The 2026 NPC & CPPCC annual sessions, which recently concluded in Beijing, included the AI Plus Initiative in the Report on the Work of the Government for the third consecutive year and, for the first time, set a goal of “creating new forms of smart economy”. Today, as rapidly advancing AI leads modern technological development and opens a new chapter in human history, China’s outstanding traditional culture, including Tibetan culture, will follow the trend and move from offline to online, from static to dynamic, and from history toward the future.

The development of DeepZang provides a valuable exploratory model for the global application of AI in processing low-resource languages—particularly Tibetan—demonstrating that modern information technology can effectively support the preservation and development of traditional culture.

DeepZang was independently developed by Xizang CHOKNOR Digital Industry Co., Ltd. Its founder, Tenzin Norbu, said at the launch event: “AI is reshaping the global information ecosystem. Every technological leap changes the way we acquire knowledge and exchange ideas. If ethnic languages are absent from this intelligent revolution, it means a rupture in cultural heritage and a deepening of the digital divide.” Spending four years, CHOKNOR has built a high-quality Tibetan-Chinese parallel corpus of nearly 70 million entries and simultaneously completed large-scale voice data collection across the three major Tibetan dialect regions, making it the largest and accurately annotated Tibetan-language speech database in China to date.

At this launch event, the DeepZang open-source large model platform—the CHOKNOR Jinyun AI Open Platform—was unveiled. This is the first domestic AI open platform for ethnic languages oriented toward multilingualism and multimodality, supporting services in more than 80 languages, including Tibetan, Chinese, English, Mongolian, Uyghur, and Kazakh, and integrating five capabilities: listening, speaking, translation, visual understanding, and reasoning. The DeepZang App was also released at the event, supporting intelligent interaction in three languages: Tibetan, Chinese, and English. Simply by speaking or typing a sentence, users can achieve real-time mutual translation, Tibetan-language question answering, cultural knowledge queries, and more.