China's DeepSeek Teams Up With Tsinghua University To Raise AI Bar, Boost Reasoning Capabilities

China's DeepSeek, in collaboration with researchers from Tsinghua University, developed a technique to improve the reasoning capabilities of large language models (LLMs) that combines generative reward modeling (GRM) and self-principled critique tuning, SCMP reported, citing a paper published on Friday.

The dual approach aims to enable LLMs to deliver better and faster results to general queries.

Reportedly, the DeepSeek-GRM models outperformed existing methods, according to SCMP, who cited the researchers.

Also Read: Rising Tide Of AI Is So Strong That Semiconductor Optical Industry Projects $30 Billion Total Addressable Market By 2029

DeepSeek aimed to make the GRM models open source. The emergence of DeepSeek and claims of affordable AI models fueled a $1 trillion market wipeout in the U.S. and a domestic price war, prompting Chinese Big Tech companies to roll out affordable AI models.

In March, DeepSeek said its upgraded V3 model offered enhanced reasoning capabilities, optimized front-end web development, and upgraded Chinese writing proficiency.

In February, it also open-sourced five of its code repositories. In late February, DeepSeek founder Liang Wenfeng participated in a symposium with tech entrepreneurs hosted by Chinese President Xi Jinping in Beijing.

Chinese e-commerce juggernaut Alibaba Group Holding BABA plans to release an upgraded version of its flagship AI model by April.

DeepSeek's claims prompted China's tech leaders to flood the market with affordable AI services.

OpenAI, Alphabet Inc GOOG GOOGL Google and Anthropic have similarly released new models.

Meta Platforms Inc META announced the release of its new Llama 4 artificial intelligence models, built on one of the world's most advanced large language models as per the company.

It is noteworthy that iShares China Large-Cap ETF FXI has gained 10% year-to-date, while iShares China Large-Cap ETF QQQ lost over 17%.

Read Next: