The development of artificial intelligence in China is expanding beyond dominant languages like Mandarin and English, with the debut of a new AI model designed specifically for the Tibetan language. According to a report by China Daily, the “Zeta” AI model has been officially launched, marking a significant milestone in the digital inclusion of minority languages. The model is expected to dramatically improve digital access, enhance public services, and facilitate cultural preservation across Tibetan-speaking regions in China.
Developed by the State Key Laboratory of Tibetan Intelligence at Qinghai Normal University, Zeta has been filed under national regulations for generative AI services and entered public testing and pilot application, making it Qinghai’s first large-scale model designed for full multimodal processing of the Tibetan language. Dorlha, the lab’s executive deputy director, described the regulatory filing as “a new starting point for applying the technology to real-world needs.” The model covers the three main Tibetan dialects: Amdo, U-Tsang, and Kham. It is built on a dataset comprising a 150-gigabyte Tibetan text corpus, 87 million multilingual parallel sentences, and 30,000 hours of multi-dialect audio recordings. It is available in three sizes: 7 billion, 50 billion, and 122 billion parameters.
Expanding Public Services and Digital Access
According to China Daily, the primary application of the Zeta model is to improve public services and bridge the digital divide. In many remote Tibetan-speaking areas, language barriers can hinder access to government services, healthcare information, and educational resources. By integrating the Zeta model into public service platforms, local governments can offer real-time translation, automated customer support, and accessible digital interfaces in native Tibetan. This capability is expected to significantly streamline administrative processes and improve the quality of life for residents who may not be fluent in Mandarin.
Furthermore, the model holds substantial promise for the education sector. AI-powered tutoring systems and educational content generators utilizing Zeta can provide personalized learning experiences for Tibetan students, helping to preserve the language while simultaneously teaching modern digital literacy skills. The initiative aligns with broader national policies aimed at integrating AI into the educational infrastructure across all regions of China.
(Related: China’s AI Plus Education Action Plan Adds AI to Teacher Qualification Exams)
Cultural Preservation in the AI Era
Beyond practical utility, the Zeta model serves as a vital tool for cultural preservation. As digital communication increasingly dominates daily life, languages that lack robust digital representation risk marginalization. By bringing Tibetan into the realm of advanced generative AI, the developers ensure the language remains vibrant and relevant in the modern technological landscape. The model can assist in translating ancient texts, generating new Tibetan literature, and facilitating cross-cultural communication.
The launch of Zeta also comes about a month after Deep-Zang, the first Tibetan large language model launched in the Xizang autonomous region (the official name of Tibet in China), offering users a second option. The emergence of two competing Tibetan-language AI models within weeks of each other highlights a critical, often overlooked aspect of China’s AI strategy: the use of technology to manage and integrate its diverse linguistic and cultural landscape. As foundational models continue to advance, the development of specialized systems like Zeta will be essential for ensuring that the benefits of the AI revolution are distributed equitably across all segments of society.