ISLAMABAD: Chinese AI startup DeepSeek has released a new version of its R1 reasoning model, boosting competition with major American AI developers.
The model, named R1-0528, is now live on Hugging Face, although no official public statement has been made.
Despite this, early results on the LiveCodeBench benchmark show it outperforming Alibaba’s Qwen-3 and xAI’s Grok-3 mini.
It’s also ranked just below OpenAI’s o4-mini and o3 in code generation performance.
Benchmark Performance and Global Response
The update is labeled a “minor trial upgrade” by the company, and developers have already started testing it.
DeepSeek first stunned the industry in January with its original R1 model, challenging the belief that only massive computing power can create top-performing AI.
That launch impacted global markets and forced competitors like Google and OpenAI to drop prices and introduce lightweight models.
China’s AI Push and Future Outlook
DeepSeek’s growth highlights China’s rapid progress in AI despite U.S. export restrictions.
The company has already released an upgraded version of its V3 LLM and is expected to launch a full R2 successor soon.
With tech giants like Tencent and Alibaba now developing competing models, China’s AI sector is witnessing explosive momentum.
As global benchmarks show R1-0528 closing the gap with U.S. leaders, the AI race continues to intensify.
Observers now await how OpenAI and others will respond to DeepSeek’s next move.
