Chinese AI startup DeepSeek has stirred the global tech scene again by quietly launching an upgraded version of its R1 reasoning model. The updated model, named R1-0528, appeared without fanfare on Hugging Face but quickly climbed the LiveCodeBench leaderboard, outperforming models from xAI and Alibaba.
Though labeled a “minor trial upgrade,” R1-0528 ranked just behind OpenAI’s o4 mini and o3 in code generation, reinforcing DeepSeek’s growing reputation in efficient AI development. This move reignites the China-U.S. AI rivalry.
Also Read: Microsoft Bans DeepSeek App for Employees Over Data Security Concerns
DeepSeek made headlines earlier in January with its original R1 model, which matched top-tier U.S. models using fewer resources. The model’s success challenged the idea that massive infrastructure is needed to scale powerful AI.
The industry now waits for the official release of DeepSeek’s R2, which was initially set for May. With Chinese tech firms like Tencent and Alibaba also pushing boundaries, global AI competition is heating up fast.
