DeepSeek has launched an urgent hiring campaign for positions related to “product and design” in Beijing and Hangzhou, sparking speculation about a new AI model the Chinese startup has yet to publicly detail.

Attracting talents for Next-Gen AI product

According to its official WeChat account, the company is hiring candidates to help build the “next-generation intelligent product experience” based on large language models (LLMs) — the foundational technology for generative AI services like ChatGPT and DeepSeek’s own chatbot.

Founded by tech entrepreneur Liang Wenfeng in 2023, this is DeepSeek’s first time recruiting roles such as product managers, product designers, and visual designers. Previously, the Hangzhou-based company had primarily focused on foundational AI model research.

Transitioning to a professional corporate structure

This recruitment drive signals that DeepSeek is evolving into a fully-fledged technology enterprise with a well-structured and professional operational model. The company is also currently hiring for key executive roles such as CFO and COO, as well as four other research and engineering positions, according to its career page.

Developing a new language model

Despite drawing significant attention from the public and investors, DeepSeek has remained tight-lipped, only releasing occasional product updates and a few research publications. The most recent upgrade to its LLM was on March 25, when it launched the enhanced open-source model DeepSeek-V3-0324 on Hugging Face.

Superior performance and cost efficiency

DeepSeek-V3-0324 has shown significant improvements in reasoning and programming capabilities over its predecessors. Standardized tests published on Hugging Face indicate enhanced performance across multiple technical benchmarks.

According to Reuters, DeepSeek may release a new reasoning model, R2, in early May, though the company has not publicly confirmed any timeline.

Global attention and impact

DeepSeek gained international attention between December 2024 and January 2025 by releasing two advanced open-source AI models, V3 and R1, developed at a fraction of the cost of major tech firms’ LLM projects.

In a blog post, DeepSeek disclosed that training V3 took only 2,048 Nvidia H800 GPUs over two months, consuming 2.8 million GPU hours at a cost of $5.6 million—a fraction of the resources typically spent by U.S. firms on similar models.

Many Chinese tech companies quickly built products based on DeepSeek’s open-source models, especially R1, released in January.

AI solutions for enterprise customers

Kai-Fu Lee, founder and CEO of Chinese AI startup 01.AI, recently stated that his company is leveraging DeepSeek’s technology to deliver enterprise AI solutions in sectors such as finance, gaming, and law. 01.AI has even halted its own model development to focus on DeepSeek’s technology.

Global watch on DeepSeek’s move

Researchers, investors, and the public are closely monitoring how DeepSeek might continue to innovate amid tightened U.S. export restrictions on Nvidia’s advanced AI chips to China.

According to the Financial Times, Nvidia CEO Jensen Huang reportedly met with DeepSeek founder Liang Wenfeng during a recent visit to China, although neither company has confirmed this meeting.

DeepSeek narrows China’s AI gap with the U.S.

Kai-Fu Lee, a prominent global AI figure and former president of Google China, told Reuters that DeepSeek has helped China surpass the U.S. in some infrastructure software engineering areas.

He believes China has closed the AI development gap with the U.S. to just three months in certain fields, thanks to companies like DeepSeek optimizing chip usage and deploying more efficient algorithms.

“DeepSeek’s ability to discover new chains of thought and improve reinforcement learning shows they are catching up quickly with the U.S. — and perhaps even out-innovating it,” Lee said, referring to the model’s ability to display reasoning processes before giving answers, a feature once pioneered by OpenAI.

01.AI shifts focus to applications

Kai-Fu Lee founded Sinovation Ventures and launched 01.AI in March 2023, joining emerging AI companies like ZhipuAI and Moonshot, as well as tech giants Baidu, Alibaba, and ByteDance in the race to build foundational AI models.

01.AI focuses on practical AI applications—software solutions that help clients deploy foundational models efficiently. In January, it launched Wanzhi, a new software platform supporting enterprise AI adoption. The company has already begun generating revenue and is projected to grow significantly in 2025, targeting multiples of the $15 million revenue it earned in 2024, according to Kai-Fu Lee.

Source: Sơn Vân (Tạp chí Một thế giới)