Chinese New Year has started on 17 February 2026, with people all around the world joining in the festivities and China celebrating Spring Festival. Several Chinese companies have launched new AI models in the lead-up to the holiday, a little over a year after DeepSeek’s now-famous V3 and R1 models.
At the end of 2024, Chinese company DeepSeek launched its much-talked-about V3 AI model, followed by its R1 model at the start of 2025. As tech news site The Information believes the next-generation V4 model will soon hit the market, rival companies have launched their new developments en masse.
Some of the most important launches include those of Alibaba, ByteDance, and Zhipu. China is generally considered to be the most important competitor to the U.S. concerning AI models, as DeepSeek even became the top-rated free application available on Apple’s App Store in the U.S. after V3 hit the market, overtaking ChatGPT.
Qwen Intelligence Files: All-in-One Vision. 🕵️♂️ Alibaba's Qwen-3.5 Plus is here! With native multimodality, it reads, hears, and sees all at once. Unified intelligence for any case. ✨
— Alibaba Cloud (@alibaba_cloud) February 19, 2026
Coding plan now just $5/mo.
🔗:https://t.co/XDafsi6Bhx#AlibabaCloud #Qwen #Qwen35Plus… pic.twitter.com/gpIdB48ieA
Alibaba: Qwen 3.5
Qwen 3.5 hit the market on 16 February 2026, and according to the e-commerce company, it can understand text, images, and videos across 200 languages. Moreover, it should be able to deploy AI agents – indecent decision-making and task-executing systems – five times faster than its competitors, while being 60% cheaper than Qwen 2.5.
“It demonstrates outstanding results across a full range of benchmark evaluations, including reasoning, coding, agent capabilities, and multimodal understanding, empowering developers and enterprises to achieve significantly greater productivity”, the company said in a press release.
Qwen 3.5 is available both in an open-weight and a hosted version.
🚀 Qwen3.5-397B-A17B is here: The first open-weight model in the Qwen3.5 series.
— Qwen (@Alibaba_Qwen) February 16, 2026
🖼️Native multimodal. Trained for real-world agents.
✨Powered by hybrid linear attention + sparse MoE and large-scale RL environment scaling.
⚡8.6x–19.0x decoding throughput vs Qwen3-Max
🌍201… pic.twitter.com/Pq0qIk54MB
ByteDance: Doubao 2.0 and SeeDance 2.0
ByteDance, the company behind social media platform TikTok, did not release one but two new AI models. Doubao 2.0 marks the second version of its chatbot and should be able to perform similarly to OpenAI’s ChatGPT and Google’s Gemini.
Aside from that, ByteDance also released the second version of its image-to-video and text-to-video app, SeeDance 2.0. Users can easily create immersive audio and video. However, the American Motion Picture Association already released a statement criticising the AI model.
“In a single day, the Chinese AI service Seedance 2.0 has engaged in unauthorised use of U.S. copyrighted works on a massive scale. 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. ByteDance should immediately cease its infringing activity”, the association said. ByteDance told the BBC it would be strengthening the current safeguards to protect intellectual copyright.
Zhipu: GLM-5
Zhipu’s open-source GLM-5 model is said to show better coding capabilities and an increased ability to perform long-running agent tasks compared to its predecessor. GLM-5 is engineered for agentic intelligence, advanced multi-step reasoning, and frontier-level performance across coding, creative writing, and complex problem-solving, according to the company.
When Zhipu went public on the Hong Kong stock exchange in January 2026, it raised HKD 4.35 billion (€465 million), which will be put into the further development of GLM-5. Due to its popularity and an insufficient number of chip clusters, however, some GLM-5 users have been experiencing some delays.
Introducing GLM-5: From Vibe Coding to Agentic Engineering
— Z.ai (@Zai_org) February 11, 2026
GLM-5 is built for complex systems engineering and long-horizon agentic tasks. Compared to GLM-4.5, it scales from 355B params (32B active) to 744B (40B active), with pre-training data growing from 23T to 28.5T tokens.… pic.twitter.com/uGYQUjIbbs
Other Chinese companies to have launched their AI models in the days before the Chinese New Year are MiniMax, NetEase Youdao, Tencent, Dexmal, and iFlytek.
