Beijing unveils Tibetan-language AI model that supports three dialects

A Tibetan-language artificial intelligence model was unveiled in Beijing on Wednesday, pushing forward China's efforts to develop AI for ethnic languages.

Named Zeta, the model was developed by the State Key Laboratory of Tibetan Intelligence under Qinghai Normal University. Its release comes about a month after the debut of DeepZang, the first Tibetan large language model launched in the Xizang autonomous region.

Built on a domestically developed computing architecture, Zeta supports Chinese AI chips and distributed inference systems, strengthening technological self-reliance. The laboratory, jointly established by Qinghai Normal University, Xizang University, and Nanjing University of Posts and Telecommunications, is the country's only national lab focused on ethnic language intelligence.

According to Dorlha, the lab's executive deputy director, Zeta is a multimodal system integrating text, speech, and visual processing. It supports three Tibetan dialects and can process complex multilingual documents, enabling functions across listening, speaking, reading, writing, and translation.

The system features a "one core model plus eight applications" framework, covering translation, speech recognition, text recognition, AI dictionaries, literature retrieval, and Tibetan medicine-related services, he said.

Backed by a large multilingual dataset, the model is expected to support digital development in sectors such as public services, education, and healthcare in Tibetan-speaking regions.