China’s artificial intelligence sector is entering the Lunar New Year with a wave of major product launches from leading tech companies, positioning the country as a significant competitor in the global AI race. Alibaba, ByteDance, and Zhipu have all unveiled new AI models in the weeks leading up to the holiday, while industry watchers anticipate a forthcoming release from DeepSeek.
Alibaba launched its Qwen 3.5 model on February 16, claiming capabilities in understanding text, images, and videos across 200 languages. The company says the model can deploy AI agents up to five times faster than previous versions, including rivals such as OpenAI’s ChatGPT and Anthropic’s Claude. The AI agents are designed to complete multi-step workflows, navigate websites, fill out forms, and even generate functional 3D games and medical image analyses. Alibaba also stated that Qwen 3.5 operates up to 60 percent cheaper than its predecessor, Qwen 2.5. The company has committed 380 billion yuan (€50.6 billion) over the next three years to cloud computing and AI development.
ByteDance introduced two new AI products in the lead-up to the holiday. The latest iteration of its chatbot, Doubao 2.0, supports complex reasoning and multi-step task execution comparable to ChatGPT and Google’s Gemini. ByteDance also rolled out SeeDance 2.0 on February 14, an app for creating audio-visual content from text and images. The app has drawn criticism from the American Motion Picture Association, which warned that it allows large-scale use of copyrighted works without adequate safeguards. ByteDance said it will enhance protections to prevent copyright violations.
Zhipu AI released GLM-5 on February 11, an open-source model designed for “agentic intelligence” with capabilities in coding, creative writing, and problem-solving. GLM-5 supports long-session workflows, full-length report generation, and advanced reasoning over academic papers. The company highlighted that the model uses DeepSeek’s sparse attention mechanism to reduce computational costs and was trained entirely on Huawei Ascend chips, achieving independence from US semiconductor hardware. Zhipu raised HKD 4.35 billion (€465 million) in a recent Hong Kong IPO to fund its next-generation AI projects.
Industry observers are also closely watching DeepSeek, known for its affordable open-source models, which is expected to launch version 4 around the Lunar New Year. DeepSeek V3, the previous model, had global impact last year, briefly overtaking ChatGPT in app rankings and contributing to a $600 billion sell-off in US tech stocks. Some European countries, including Italy, Denmark, and the Czech Republic, have already banned the use of DeepSeek models in government agencies due to cybersecurity and data privacy concerns.
The flurry of AI releases in China highlights the country’s rapid advancement in autonomous AI agents capable of executing complex tasks with minimal human input. Analysts expect competition with US tech firms to intensify as these models gain adoption both domestically and internationally.
