Alibaba Previews Qwen3.8 Max, a 2.4 Trillion-Parameter Multimodal AI Model

My take: Alibaba unveiled Qwen3.8 Max at the World AI Conference in Shanghai, with 2.4 trillion parameters in a sparse Mixture-of-Experts architecture and full multimodal support: text, images, video, and documents. It is the first Qwen model above one trillion parameters to handle all of that together.

The central claim is that it ranks "second only to Fable 5," but there is an important caveat: that ranking comes from Alibaba's own internal evaluations, with no public benchmark table, no model card, and no disclosed architecture or training data details. When the company that builds the model is also the one scoring it, it is worth waiting for independent results before treating that position as definitive.

Even so, the level that Chinese open-weight AI has reached is a meaningful signal for the market. Twelve months ago, a 2.4-trillion-parameter open model was science fiction. The launch pricing, at 10% of the standard rate on Alibaba's platforms, makes it even harder to ignore.

Does your AI strategy account for Chinese open-weight models, or are you still betting exclusively on Western labs?