Alibaba's AI research unit Qwen has released details of a new large language model with 2.4 trillion parameters. The company says the model's open weights will be made available to developers and researchers next week.
What the model is
The model, which Qwen has not yet named publicly, is one of the largest language models ever disclosed by a Chinese tech firm. Its 2.4 trillion parameters put it in the same weight class as some of the biggest models from Western labs, though direct comparisons are tricky because architecture and training data differ. Qwen is part of Alibaba's broader push into generative AI, competing with offerings from Baidu, Tencent, and a growing number of Chinese startups.
Open-weight release
Unlike some competitors that keep their largest models proprietary, Qwen plans to release the model's weights under an open license. That means anyone can download the trained parameters and run the model on their own hardware, subject to the license terms. The company hasn't specified which license it will use or whether there will be restrictions on commercial use. The release is scheduled for next week, though an exact date hasn't been announced.
Open-weight releases have become a flashpoint in the AI industry. Some argue they accelerate research and democratize access; others worry about misuse, such as generating disinformation or building harmful applications. Qwen's decision to open up a model of this scale could reignite that debate. The company has previously released smaller models openly, but this is by far its largest.
Alibaba hasn't said how much it cost to train the model or what hardware it used. The company's cloud division, Alibaba Cloud, is likely to offer the model as a service, but the open weights give developers an alternative path.
Next week's release will be closely watched by the AI community. The question now is how quickly developers adopt it — and what they build with it.
