https://japanese.cri.cn/2026/03/17/ARTI1773715425482854

DeepZang, the world’s first highly responsive artificial intelligence (AI) large-scale language model (LLM) for the Tibetan language, was unveiled on March 15th in Lhasa, Tibet Autonomous Region, in southwestern China.

DeepZang was developed by Tibetan Gioro Digital Industry Management. The open-source large-scale language model is China’s first ethnic language AI open platform that supports multiple languages ​​and modalities, executing over 80 language services and realizing the functions of ‘listening, speaking, translating, seeing, and thinking.

A DeepZang app was also released simultaneously and is expected to be adopted in many industries. Gioro Digital finalized its “Tibetan-Chinese Bilingual + AI” development plan as early as 2018. Over the next four years, the company built a parallel corpus of translation samples of nearly 70 million high-quality Tibetan and Chinese texts. Simultaneously, they conducted extensive audio collection across the three major Tibetan dialect regions, recording 10,500 hours of the Utsang dialect, 10,000 hours of the Kham dialect, and 10,000 hours of the Amdo dialect, creating the largest and most accurate Tibetan audio database in China.

The World Records Certification Organization (WRCA) has awarded Deep Zhan certification for the “world’s first large-scale Tibetan language model” at the company’s launch event.