Tencent Holdings Ltd. became the latest tech company to unveil or enhance an AI model intended to eclipse DeepSeek, joining a spate of rollouts since the startup’s emergence energized the U.S.-China technology race.
The Hunyuan Turbo S artificial intelligence model is designed to respond as instantly as possible, distinguishing itself from the deep reasoning approach of DeepSeek’s eponymous chatbot. The deployment cost has also fallen sharply, Tencent said on its official WeChat channel.
From OpenAI to Alibaba Group Holding Ltd., major industry players on both sides of the Pacific have in just the past month rolled out AI models at a rapid clip. The series of introductions underscore a dramatically quickened pace of development since DeepSeek stunned Silicon Valley with a model that matched the best from OpenAI and Meta Platforms Inc. That was true particularly in China, where the 20-month-old startup has galvanized interest across a tech industry that for years struggled to match the US.
“Every player will incorporate their techniques to improve their own architecture, and so we’ll keep having companies come up with new models that claim to be better,” said Xin-Yao Ng, an investment director at abrdn plc. “Tencent is likely to also invest heavily into AI infrastructure like Baba, and look to grow their AI cloud business.”
DeepSeek, founded by quantitative hedge fund manager Liang Wenfeng, has upended the global AI sector since it burst onto the scene in January.
Almost overnight, DeepSeek dismantled many of the assumptions inside Silicon Valley about the economics of building AI, as well as the best technical methods for developing the technology and the extent of the U.S. lead over competitors in China. It sparked the biggest stock wipeout in Nvidia Corp.’s history last month, while creating hope for a renaissance for China’s leading internet companies.
The Turbo S is now available to developers and enterprise users via Tencent’s cloud services. It will begin testing the model among selected users on its consumer-facing chatbot, Yuanbao, which had lagged rivals in usage.
Tencent’s release comes a month after Alibaba benchmarked its latest Qwen AI model against DeepSeek. On Thursday, the WeChat operator shared figures showing its own Hunyuan Turbo S was competitive against DeepSeek’s V3 model in commonly used AI tests. Baidu Inc. last week said it would soon roll out its latest “Ernie” platform. And ByteDance Ltd. is testing a new model similar to DeepSeek’s R1.
Many of them are taking aim squarely at DeepSeek, which since its introduction has been widely embraced by corporate and government entities across China. This month, President Xi Jinping invited Liang to join a select forum of tech leaders including Alibaba co-founder Jack Ma, in effect designating the startup a national technology champion.
DeepSeek has proven so popular it withheld certain services for a time because of a lack of server capacity. To address surges in demand, it slashed pricing to access its application programming interface during off-peak hours.
Last week, the company promised to release key code and data to the public, an unusual step to share more of its core technology than rivals such as OpenAI have done. With that move, DeepSeek is pushing harder on an open-source approach to AI development that’s won more advocates of late, and paved the way for wider adoption of its own technology.
