Over half of China's large language models are open sourced: report

BEIJING, May 29 (Xinhua) -- China has 14 provincial-level regions that are currently working on the development of large language models (LLMs), among which more than half are open sourced, according to a report released at the ongoing Zhongguancun Forum in Beijing.

LLMs are computer programs that analyze and create text.

The report, issued by the Institute of Scientific and Technical Information of China (ISTIC), said that Beijing has the greatest number of LLMs, followed by south China's Guangdong Province.

The country's LLM research has mainly focused on the fields of natural language processing and multimodal LLMs, while other areas such as computer vision and intelligent speech have received limited attention, said Zhao Zhiyun, head of the ISTIC.

Public computing power across the country has been provided to satisfy the rapidly growing demand for artificial intelligence and LLM development computing power, according to the report.

The report also called for more global cooperation and suggested that scenario-driven LLM development should be reinforced.