These are China's new AI models that have just been released ahead of the Lunar New Year
Major Chinese AI companies such as Alibaba, ByteDance, and Zhipu have all announced launches in the weeks leading up to the Lunar New Year, while the industry awaits a possible new drop byDeepSeek.
China is ringing in the Lunar New Year with a flurry of new artificial intelligence (AI) model launches. Tech companies, such as Alibaba, ByteDance, and Zhipu, have all announced new product launches in the weeks leading up to China’s biggest holiday, while industry watchers expect a new Deepseek model soon.
China is widely regarded as a major competitor to the United States in the race to adopt and develop artificial intelligence models.
Thenew deployments are preparing the Chinese market for AI agents, systems that can make decisions and execute tasks, such as navigating websites and generating content, without needing human assistance.
Alibaba’s Qwen 3.5
E-commerce giant Alibaba released its latest AI model, Qwen3.5, hours before the Lunar New Year starts on 16 February. The model understands text, images, and videos across 200 languages, the company says.
The new model can deploy AI agents up to five times faster than previous models and its competitors, including the latest models of OpenAI’s ChatGPT and Anthropic’s Claude, the company said. Their agents can fill out forms, navigate websites, and complete multi-step workflows.
Early tests show Qwen3.5 can generate functional 3D games, browsers, websites, and analyse medical imagery. The model is also up to 60 percent cheaper than its predecessor model, Qwen2.5, the company said.
In 2025, Alibaba committed 380 billion yuan (€50.6 billion) to cloud computing and AI in the next three years, one of the company’s largest tech investments to date.
ByteDance’s launches
ByteDance, the Chinese company behind TikTok, announced two new AI developments in the weeks before the Lunar New Year.
The latest version of the company’s AI chatbot, Doubao 2.0, launched over the weekend. The newest model includes complex reasoning and multi-step task execution that matches OpenAI’s ChatGPT and Google’s Gemini current models, Reuters reported.
ByteDance also released SeeDance 2.0 on February 14, the second version of its image-to-video and text-to-video app. The software lets users create “immersive” audio and video with director-level controls. One viral video features a rooftop fight between American actors Tom Cruise and Brad Pitt.
The American Motion Picture Association criticised SeeDance2.0 for unleashing the use of copyrighted works on “a massive scale,” according to a statement.
“By launching a service that operates without meaningful safeguards against infringement, ByteDance is disregarding well-established copyright law that protects the rights of creators and underpins millions of American jobs,” the statement reads.
ByteDance representatives told the BBC that it will take steps to strengthen current safeguards on the platform to protect intellectual copyright..
Zhipu AI’s GLM-5
Zhipu AI released its open source GLM-5 model on February 11, a model that the company says is engineered for “agentic intelligence, advanced multi-step reasoning, and frontier-level performance,” in coding, creative writing and problem-solving.
The model can build assistants that plan, browse, call tools and manage multi-step workflows over long sessions, the company says. It can also generate full-length reports and process and reason long academic papers.
GLM-5 uses DeepSeek’s sparse attention mechanism DSA, which cuts computational costs while enhancing model efficiency.
The company claims the model was also trained entirely on Huawei Ascend chips and “achieves full independence from US-manufactured semiconductor hardware,” which the company said is a “milestone in self-reliant AI infrastructure”.
Last month, Zhipu went public on the Hong Kong stock exchange and raised HKD 4.35 billion (€465 million) for what the company said will be its next-generation model development.
Coming soon: DeepSeek’s V4
DeepSeek, the Chinese AI company known for its open source cheap models, is expected to release its fourth version sometime around the Lunar New Year, according to The Information.
When V4 is released, it could replace the V3 model, which powered the assistant app that became the top-rated free application available in the United States and worldwide, overtaking ChatGPT last January.
DeepSeek’s V3 rattled global markets last year and sparked a global sell-off of US-led tech stocks. Industry giants, including chipmaker Nvidia, saw shares plummet by 17 per cent and erased $600 billion (€573 billion) in market capitalisation before recovering later in the day.
While it hasn’t been released yet, DeepSeek fuelled anticipation last week when its chatbot upgraded its context window, the amount of information that it can remember and handle in a single task, according to the South China Morning Post.
Some European countries, such as Italy, Denmark, and the Czech Republic banned government agencies from using DeepSeek models on their devices due to data security and cybersecurity concerns. Local media in Belgium reported that government officials stopped using DeepSeek in December.
No comments:
Post a Comment