BEIJING, China: A year after DeepSeek jolted global markets with a low-cost artificial intelligence model, Chinese technology firms are preparing a new wave of releases, many timed around the country's busiest holiday season.

The Hangzhou-based startup's breakout during the 2025 Spring Festival reshaped China's AI landscape, pushing low-cost, open-source systems to the centre of the industry. This year, several rivals are expected to unveil upgraded models as the holiday period begins on February 15.

When DeepSeek first launched a powerful model despite U.S. export controls limiting access to advanced semiconductors, the industry was caught off guard. Attention has now shifted to what Chinese developers will deliver next.

"The surprise would be if some of these new models end up being underwhelming. I think there are high expectations here," said Alfredo Montufar-Helu, managing director at Ankura Consulting in Beijing. Zhipu AI released a new model, it said, with stronger coding capabilities and the ability to handle long-running tasks without prompts. ByteDance followed on February 12 with Seedance 2.0, a video-generation system described by the state-backed Global Times as "capable of producing cinematic blockbusters in seconds".

ByteDance is also expected to upgrade its Doubao chatbot, which is currently China's most popular AI app, with 155.2 million weekly active users, according to QuestMobile.

DeepSeek is preparing its next-generation V4 model, while Alibaba is expected to unveil its Qwen 3.5 series with enhanced mathematical reasoning and coding capabilities, according to The Information. Developers for Qwen submitted support code to the open-source platform Hugging Face this month, often signalling an imminent release. The companies have not announced formal launch dates.

DeepSeek's initial release in January 2025 triggered a global tech selloff, wiping US$593 billion from Nvidia's market value in one day. Its models have consistently undercut competitors on price, lowering usage costs below many U.S. offerings.

U.S. investors saw DeepSeek's claim to have built a model comparable to OpenAI's best as a challenge to the idea that only firms spending tens of billions on computing infrastructure could create advanced AI. A RAND report last month found Chinese models operate at roughly one-sixth to one-fourth the cost of comparable U.S. systems.

"DeepSeek showed the industry that you can create a very good model even when you're resource-constrained," said Lian Jye Su, chief analyst at Omdia.

"The combination of open-source access, strong reasoning capabilities, and low deployment costs has become a defining model for how Chinese vendors now approach foundation models."

Since DeepSeek's breakthrough, open source has become mainstream among Chinese developers. Baidu and other firms began releasing portions of their own models after DeepSeek briefly overtook ChatGPT in U.S. app downloads. Hugging Face is now dominated by releases from Baidu, ByteDance, Tencent, and startups such as Moonshot.

"Chinese companies are actively embracing open source, significantly lowering the barriers for global developers and enterprises to access cutting-edge AI technology," Global Times wrote.

While DeepSeek remains focused on improving core model performance, rivals are increasingly integrating AI into consumer services. Alibaba's Qwen chatbot has tested, allowing users to buy goods directly via chat.

DeepSeek differs structurally from competitors. Its parent is a quantitative hedge fund controlled by founder Liang Wenfeng, enabling it to prioritise research over commercial returns and operate without outside shareholder pressure.