Alibaba Group has previewed Qwen3.8, the next-generation frontier model in its Qwen family, and says it is "second only" to Anthropic's Claude Fable 5 among widely released systems. The announcement, made on Sunday during the World Artificial Intelligence Conference in Shanghai, marks the first time a Chinese lab has publicly claimed the world's number-two spot.
2.4 trillion parameters, and a paid preview already live
Qwen3.8-Max-Preview, the preview version of the strongest model in the family, is already routable via Alibaba's Token Plan subscription and its Qoder and QoderWork agentic platforms, the company said in an official X post on Sunday. With 2.4 trillion parameters, Alibaba called Qwen3.8 "one of the most powerful models available today, comparable to leading frontier AI models," trailing only Fable 5.
China enters the multi-trillion-parameter age
The preview lands a week after Beijing-based Moonshot AI unveiled Kimi K3 with 2.8 trillion parameters, currently the largest open-source AI model. Zhipu is expected to counter with GLM-5.2 at WAIC, and DeepSeek is rumored to be readying its own next-generation update. Taken together, the moves push Chinese labs into direct scale competition with US heavyweights OpenAI, Anthropic and Google.
WAIC as a Chinese AI showcase
The Qwen3.8 preview lands at a WAIC that has become a coordinated Chinese answer to Washington's export controls. Xi Jinping used the opening day to launch the World AI Cooperation Organisation with 29 signatory countries, while dozens of Chinese labs have used the show to release new humanoids, world models and chips. AgiBot debuted four humanoids, and Shanghai's ACE ROBOTICS unveiled Kairos 3.1, a world model built for embodied AI.
Reporting based on coverage from the South China Morning Post and Bloomberg.