- China Telecom’s AI Institute trained a 1-trillion-parameter LLM using tens of thousands of Chinese-made chips.
- The achievement underscores China’s progress in developing advanced AI capabilities independent of foreign technology.
What happened: China Telecom achieves AI milestone with domestically trained 1-trillion-parameter model
China Telecom‘s Institute of AI announced the successful training of a large language model (LLM) comprising one trillion parameters, utilizing tens of thousands of domestically produced chips. This development represents a significant advancement in China’s efforts to achieve technological self-sufficiency in artificial intelligence, particularly in the face of tightening U.S. export controls on advanced semiconductors.
The company also introduced an open-source model, TeleChat2-115B, trained on the same domestic hardware. While specific details about the chips used were not disclosed, China Telecom has previously collaborated with Huawei, suggesting the possible use of Huawei’s Ascend series AI processors.
Also read: Loxley Wireless and China Telecom sign MoU to partner with Thailand’s TOT
Also read: US state targets local telecom firms over China links
Why it’s important
This accomplishment highlights China’s growing capabilities in developing advanced AI technologies without reliance on foreign hardware. By successfully training a trillion-parameter model using domestically produced chips, China Telecom demonstrates the country’s progress in overcoming challenges posed by international trade restrictions.
The development also reflects a broader trend of Chinese tech companies investing in homegrown solutions to reduce dependence on imported technology. As the global landscape of AI continues to evolve, such advancements position China as a formidable player in the field, with the potential to influence international standards and practices in artificial intelligence.
