On January 27th, a remarkable event transpired overnight that sent ripples throughout the global tech communityThe "mysterious power from the East," as it has been duly noted, made its presence felt among overseas enthusiasts, thanks to the emergence of DeepSeek, a rising star in the AI landscape.

In the past couple of days, this domestic AI powerhouse has taken the world by stormFollowing the launch of its new model, a surge in traffic nearly caused DeepSeek's servers to crash, reminiscent of the previous AI sensation when Kimi from the "Dark Side of the Moon" became a household nameFortunately, the issue was swiftly addressed within minutes.

As of the time this article is being written, DeepSeek's app has climbed to the second spot in the free rankings of Apple's App Store in the United States, right behind ChatGPTThis unprecedented rise underscores the growing impact of this Chinese AI entity on the international stage.

What triggered all this commotion? It was the formal unveiling of DeepSeek-R1, its inference model released on January 20thThe model has garnered attention for its remarkable performance in critical areas such as mathematics, programming, and reasoning, positioning itself as a competitor capable of matching OpenAI’s strongest inference model, GPT-4. However, here’s the kicker: while OpenAI commands premium prices, DeepSeek's API calls are remarkably more economical, costing a staggering 90-95% less.

The launch of DeepSeek-R1 sparked significant interest from the international marketProminent figures in the AI sector in the United States lauded the model as a game changerNotably, Andrew Ng, a renowned computer scientist and Sam Altman's mentor, expressed his admiration at the 55th World Economic Forum in Davos, stating, “I am impressed by the advancements of DeepSeekThey have managed to train models in a very cost-effective manner, and the performance of their newly released inference model is exceptional… ‘Keep pushing forward!’”

Sundar Pichai, the CEO of Microsoft, also publicly acknowledged DeepSeek’s contributions

Advertisements

He remarked, “They (DeepSeek) have genuinely developed an open-source model that excels in inference computation and exhibits remarkable supercomputing efficiency.” He emphasized the importance of taking China’s advancements in technology seriously, noting the groundswell of innovation emerging from the region.

Founded in May 2023, DeepSeek sprang from the vision of Huansheng Quantitative, a hedge fund giant in ChinaThe company has since made rapid strides in the AI sector, launching its first model, DeepSeek Coder, on November 2, 2023. This model is noteworthy not only for being free for commercial use but also for being completely open-sourceBy November 29, DeepSeek had expanded its offerings, releasing the DeepSeek LLM with an impressive 67 billion parameters, closely rivaling GPT-4, alongside a chat version called DeepSeek Chat.

However, it was the open-source release of the next-generation MoE model, DeepSeek-V2, in May 2024 that truly catapulted DeepSeek into the global spotlightWith performance that can rival GPT-4 Turbo yet priced at only a fraction—one percent—of the cost, DeepSeek earned the nickname “price butcher” and was dubbed “the Pinduoduo of the AI realm.” Pinduoduo is renowned for its disruptive, low-cost model in the e-commerce space, emphasizing the significance of affordability in technology.

The latter half of 2024 was marked by further innovations from the company, with the introduction of DeepSeek R1-lite-preview and DeepSeek-V3. By 2025, the R1 model demonstrated astonishing results in mathematical aptitude tests, achieving a 77.5% accuracy rate on the MATH benchmark, a performance comparable to OpenAI's outputMoreover, in programming, R1 scored a stellar 2441 on Codeforces evaluations, surpassing 96.3% of human participants.

Remarkably, all of this was accomplished with an investment of under six million dollars and the utilization of 2048 low-performance H800 chips, with a training duration of only two months

Advertisements

This “minimal input for maximal output” approach has fundamentally challenged the notion that substantial funding and resources are prerequisites for success in AI—an understanding that has taken the world by surprise.

In an era filled with competition among seven leading large model startups in China, DeepSeek has maintained a notably low profileWhile other high-profile companies invest heavily in advertising and brand marketing, DeepSeek has notably operated without a public relations team, a fact unknown to many until now.

In April 2023, Huansheng Quantitative announced the formation of a new entity devoted to exploring the essence of AGIThe company made it clear that, for years, it had consistently funneled a significant portion of its revenues into the field of AI, establishing leading-edge hardware infrastructure and undertaking large-scale research endeavors aimed at uncovering the unknown mysteries of humanity.

Reflecting on the over one year journey, it becomes clear why DeepSeek has managed to unsettle Silicon Valley in the ongoing battle for AI supremacyConversations among AI professionals and investors in WeChat circles have highlighted that the company's technical strength is just the tip of the iceberg; it is the innovative mindset and the talented workforce that set DeepSeek apart from its competitors.

Founded by Liang Wenfeng, a graduate of Zhejiang University with a background in information and communication engineering, DeepSeek has garnered a stellar reputation in the tech industry for its founder's profound technical idealism and dedication to innovationLiang continues to maintain a low profile, immersing himself in research alongside frontline researchers by reading academic papers, writing code, and participating in group discussions on a daily basis.

He stated in an interview, “For many years, Chinese companies have relied on others for technological innovation, merely adapting these innovations for their own applications

Advertisements

This is not a givenOur goal this time around is to not just take advantage of opportunities but to position ourselves at the cutting edge of technology and drive the entire ecosystem forward."

In its recruitment announcements, DeepSeek boldly proclaims "hiring top-tier talent." According to publicly available information, the company’s current team consists of a significant number of young talents graduated from renowned Chinese universities, including fresh graduates and internsExperience is no longer the sole criterion for talent selectionDeepSeek's human resources have noted on social media that they place a premium on candidates' potential and passion for large models.

The company ensures that any promising technical proposals by employees receive full support in terms of computing power and resourcesIn the sphere of large models, where computing capability is scarce, DeepSeek sets itself apart by offering “large GPU training clusters without application requirements, allowing unrestricted usage.”

Those familiar with DeepSeek report that the company provides highly competitive remuneration to its talentsIn an environment where innovation and curiosity are encouraged, DeepSeek’s mission is emblematic of a broader narrative surrounding "hardcore technology innovation" emerging from China.

In recent developments, Yushu Technology unveiled an impressive demonstration video showcasing its latest B2-W robotic dog, which showcased a series of challenging maneuvers that sparked Elon Musk's admiration and went viralCoincidentally, during Nvidia CEO Jensen Huang’s conference in China, Yushu’s CEO, Wang Xingxing, was invited to demonstrate alongside other notable figures in the AI field.

These emerging innovators and their cutting-edge companies are leading the charge in a new chapter of China’s innovation story, further establishing the country’s position as a critical player in the global technology landscape.

Advertisements

Advertisements