.png){kind=logo}
Chinese AI startup DeepSeek has quietly slipped in an update to its R1 reasoning model. This update occurred as a competitive move against US AI developers like OpenAI. The new update, R1-0528, silently appeared on the messaging platform Hugging Face without an official announcement or a hazy description of its functionalities.
However, the benchmark results show that DeepSeek's model has reached a close second to OpenAI's finest, being the o4 mini and o3 reasoning models, in code generation performances.
In accordance with LiveCodeBench leaderboard records compiled by a team of researchers from UC Berkeley, MIT, and Cornell, DeepSeek's R1-0528 falls just behind the best OpenAI models, having outperformed its rivals, such as Grok 3 mini by xAI and Qwen 3 by Alibaba, in some tasks.
This update again restores DeepSeek's claim as a reputable contender in AI, narrowing the gap with U.S.-based AI giants in reasoning and code generation capability.
DeepSeek challenged the prevailing view that China’s AI development was limited by U.S. export controls. They sent ripples through the industry by releasing AI models that rivaled, if not outperformed, top American models while requiring only a fraction of the computation and cost.
Such developments shook up the global tech markets and compelled Chinese tech companies like Alibaba and Tencent to fast-track their own AI development efforts.
Presently, the much-anticipated next-generation R2 holds all the excitement in the market, with DeepSeek's R1 already leaving an undeniable impact on the marketplace. The projected initial release in May got delayed, but interest in this new model remains high.
At the same time, elimination of price competition comes into play, with Google and OpenAI providing increasingly inexpensive models in comparison with others: discounted access tiers for Gemini from Google, o3 Mini from OpenAI, etc.
The continued developments from DeepSeek signal a growing presence of China in front-line AI technology. By making more powerful models requiring less computation, DeepSeek breaks very ancient concepts of AI supremacy.
The flow from DeepSeek also signifies that AI innovation is now increasingly globalized, and China has emerged as a strong competitor in this critical technology field.
