The world’s first Tibetan large language model, DeepZang, was launched on the 15th in Lhasa, Tibet. DeepZang was independently developed by Jue Luo Digital Industry Management Co., Ltd. in Tibet.
According to Sing Tao Daily, founder Danzeng Luobu said that this open-source large model platform is China’s first multi-language, multi-modal open AI platform for minority languages. It supports services in more than 80 languages and realizes functions such as "listening, speaking, translating, viewing, and reasoning".
Danzeng Luobu said that the R&D team spent four years building a high-quality Tibetan-Chinese parallel corpus of nearly 70 million entries. At the same time, they completed large-scale speech collection in three major Tibetan dialect areas, creating what is currently one of the most accurately labeled Tibetan speech databases.