If you’ve been following the fast-moving world of artificial intelligence, you’ll want to know about DeepSeek. This Chinese AI start-up has just launched a big upgrade to its existing reasoning model – and it’s getting attention from all corners of the globe.
On May 30, DeepSeek announced the release of R1-0528, the first major update to its original R1 model, launched in January. The company says the new version performs just as well as the world’s best models from OpenAI and Google, including OpenAI’s O3 and Google’s Gemini 2.5-Pro.
Moreover, R1-0528 has even outperformed the top AI model from Alibaba, one of China’s biggest tech companies. That’s a major achievement, especially in such a competitive space.
Stronger writing, better reasoning, and fewer mistakes
With this update, the AI’s writing and reasoning abilities improved greatly. It can now create essays, stories, and other types of creative writing that feel more natural and human. If you use AI for writing help, this makes a big difference.
Its coding skills have also been boosted, making it more helpful for developers. One of the most notable changes is a 50% drop in so-called “hallucinations” – a common problem in AI models where the system gives you wrong or made-up information.
deepseek is so back…
— Haider. (@slow_developer) May 28, 2025
the new version of deepSeek r1 is just incredible at coding
i run a coding challenge where the model builds a word scoring system
the model reasoned briefly and then gave me two files: clean code and working tests
both ran perfectly on the first try.… pic.twitter.com/HYztXPJ6Jb
These improvements didn’t come out of nowhere. DeepSeek says it puts more computing power into the post-training stage, the final phase where the model gets fine-tuned. This step helps make the AI more accurate, safer, and efficient.
According to benchmark tests, R1-0528 now leads all Chinese AI models when it comes to solving math problems, writing code, and understanding complex logic. It has also climbed the ranks to sit alongside top global names in the Artificial Analysis Intelligence Index, a well-known tool that compares AI models.
Outpacing Alibaba and taking on the world
Just a month ago, Alibaba’s model Qwen3 had overtaken the original R1 model in the LiveBench rankings for open-source AI systems. But now, DeepSeek’s latest release has taken back the top spot. That means DeepSeek is not just keeping up – it’s leading the charge.
Artificial Analysis, a respected AI research group, said DeepSeek has moved ahead of big names like Meta, Anthropic, and Elon Musk’s xAI. They’ve ranked DeepSeek as tied for second place globally among AI labs, just behind OpenAI. That’s no small feat for a start-up.
DeepSeek is aiming for the king: o3 and Gemini 2.5 Pro pic.twitter.com/XjQQNRMSEa
— Lisan al Gaib (@scaling01) May 29, 2025
The gap between closed-source models (like those from OpenAI) and open-source models (like DeepSeek’s) is also shrinking fast. This could be good news for developers, as open-source models are usually easier and cheaper to use and adapt.
Widespread support from tech giants
After the launch, DeepSeek’s upgraded model quickly gained support from many of China’s top tech companies. Tencent, Baidu, and ByteDance announced they would include R1-0528 in their cloud services. That means more developers and businesses can try to build new tools.
But the interest isn’t only local. International AI infrastructure firms like Fireworks AI and Hyperbolics are also jumping on board. The global response shows how much impact this model already has.
In addition to the full-sized model, DeepSeek also introduced a smaller version called DeepSeek-R1-0528-Qwen3-8B. Despite being almost 30 times smaller, it performs at the same level as Alibaba’s massive Qwen3-235B. That’s possible thanks to a process called “distillation,” where knowledge from a larger AI is passed down to a smaller one.
This method could help researchers and developers build faster, cheaper, and more efficient AI tools without sacrificing quality.
With its growing reputation and strong results, DeepSeek is proving it’s not just a local player – it’s becoming a global force in AI.
