Key Takeaways
- Chinese AI startup DeepSeek quietly released an upgraded version of its reasoning model.
- The new model boasts improved performance in reasoning, math, and coding.
- It ranks competitively against models from giants like OpenAI.
- This development underscores China’s continued advancements in AI despite U.S. tech restrictions.
Chinese AI firm DeepSeek has subtly launched an enhanced version of its artificial intelligence reasoning model, continuing its trend of low-key releases.
The company didn’t issue a formal announcement for the DeepSeek R1 upgrade. Instead, it appeared on Hugging Face, an AI model repository, similar to its predecessor’s debut.
DeepSeek gained attention earlier this year when its free, open-source R1 model impressed by outperforming offerings from major players like Meta and OpenAI. This rapid, cost-effective development sent ripples through global markets, raising questions about the spending habits of U.S. tech giants.
This latest model is a reasoning AI, meaning it’s designed to tackle more complex tasks by thinking through problems step-by-step, much like a logical process.
On LiveCodeBench, a platform that benchmarks AI models, the upgraded DeepSeek R1 is performing impressively, positioned just behind OpenAI’s o4-mini and o3 models, as CNBC reported.
Adina Yakefu, an AI researcher at Hugging Face, told the news outlet that DeepSeek’s latest upgrade “is sharper on reasoning, stronger on math and code, and closing in on top-tier models like Gemini and O3.”
Yakefu highlighted “major improvements in inference and hallucination reduction,” explaining that “hallucination” refers to instances where AI provides incorrect information. She added, “this version shows DeepSeek is not just catching up, it’s competing.”
DeepSeek’s progress serves as an example of how Chinese AI development is flourishing despite U.S. efforts to limit China’s access to crucial technology, including advanced chips.
This month, Chinese tech companies Baidu and Tencent shared how they are optimizing their AI models to navigate U.S. semiconductor export restrictions effectively.
Nvidia CEO Jensen Huang, whose company designs essential graphics processing units for training large AI models, recently criticized U.S. export controls.
Huang stated, “The U.S. has based its policy on the assumption that China cannot make AI chips. That assumption was always questionable, and now it’s clearly wrong.”
He further emphasized, “The question is not whether China will have AI. It already does.”
