Beijing's first factory for producing units of text for large language models (LLMs) has reached a daily production capacity of more than 1.4 trillion tokens, according to the Administrative Committee of the Beijing Economic-Technological Development Area, also known as Beijing E-Town.
The factory produces tokens, which are the smallest unit of text processed by LLMs and other artificial intelligence systems to understand and generate human language.
Recently completed in Beijing E-Town, the Beijing No 1 Token Factory integrates computing power, semiconductor chips, high-quality datasets, and LLMs to improve the token output per unit of energy consumption while enhancing overall system stability, according to a press release issued by the development area.
Unlike conventional computing centers that mainly lease raw computing resources, the token factory provides hardware infrastructure and software services simultaneously, transforming computing power into production-ready tokens that can be directly used by AI applications.
According to the development area, about 50 percent of computing tasks are completed within six seconds, while 90 percent receive responses within 10 seconds. Performance fluctuations are kept within 20 percent.
