Chinese tech rivals ready second wave of AI models on anniversary of DeepSeek disruption

One year after Chinese startup DeepSeek upended the global technology industry with its low-cost artificial intelligence model, rivals in China appear far better prepared, entering the fray with new models that offer features potentially more attractive to users.

The Hangzhou-based firm’s meteoric rise during the Lunar New Year festival in early 2025 transformed the country’s AI ecosystem, thrusting low-budget and open-source models into the spotlight.

As the primary holiday period approaches—officially commencing on February 15, 2026—a raft of companies, not limited to DeepSeek, is expected to unveil new products.

Alfredo Montufar-Helu of Ankura Consulting noted that DeepSeek shook the industry by debuting a powerful model despite US export controls restricting access to advanced semiconductors. Now, however, the market is keenly awaiting the Chinese companies’ next move.

“It would be a surprise if some of these new models fell short of expectations. I think expectations are quite high,” Montufar-Helu said.

Tech giants roll out next-generation products

On Wednesday, Zhipu AI released a new artificial intelligence model featuring enhanced coding capabilities and the capacity to execute long-duration tasks without user prompts.

On Thursday, ByteDance officially introduced Seedance 2.0, a video generation model reportedly capable of producing cinematic blockbusters in seconds.

ByteDance is also expected to present an upgraded version of Doubao, currently China’s most popular chatbot according to QuestMobile data.

According to a January report by The Information, DeepSeek is also preparing to launch its next-generation V4 model, boasting advanced aptitudes in mathematics and programming.

In early February, developers behind Qwen shared auxiliary code for the upcoming Qwen 3.5 series on the open repository Hugging Face—a move that typically signals an imminent launch.

While Alibaba, ByteDance, and DeepSeek have not announced official launch dates for their next-generation models, the companies did not respond to requests for comment on the matter.

Open-source and low-cost models become the new standard

A report on AI competition published in January by the RAND research group stated that the operating costs of Chinese models range between one-sixth and one-tenth of comparable US systems.

“DeepSeek proved to the industry that a very good model can be built even with limited resources,” said Lian Jye Su of Omdia. Su noted that the combination of open-source utilization, high reasoning capacity, and low deployment and operating costs has now become the defining approach.

Prior to DeepSeek’s breakthrough, some industry leaders, including Baidu CEO Robin Li, had asserted that the market would be defined by closed models.

However, days after the DeepSeek assistant surpassed ChatGPT in download numbers on the US app store, Baidu and other major corporations began opening up segments of their own systems.

Chinese dominance grows on Hugging Face platform

The Hugging Face platform is currently inundated with releases from tech titans such as Baidu, ByteDance, and Tencent, alongside startups like Moonshot.

A Global Times report on Wednesday stated, “Chinese companies’ active shift to open source significantly reduces barriers to access advanced AI technologies for developers and companies worldwide.”

Alongside this strategic shift, DeepSeek’s rivals have accelerated efforts to recruit leading researchers in the AI field.

Strategic divergence and commercial realities

While DeepSeek remains focused on enhancing the performance of its foundational model, competitors are prioritizing the integration of AI into consumer services.

For instance, Alibaba’s Qwen chatbot has recently begun trialing a feature that allows product purchases directly through dialogue commands.

This shift reflects commercial realities. Companies like Alibaba face shareholder pressure to monetize AI investments through consumer and enterprise products while continuing to finance infrastructure expansion.

DeepSeek’s structure, however, continues to diverge from its rivals. Being a subsidiary of a quantitative hedge fund controlled by founder Liang Wenfeng allows DeepSeek to prioritize research over commercial concerns and evade the pressure of external investors.