MANY MINDEDAlibaba Previews a Trillion-Parameter Open-Weight Model, With Weights Promised Soon
Alibaba has unveiled Qwen 3.8, an open-weight AI model with 2.4 trillion parameters. The Qwen team says it matches frontier models and trails only a system referred to as Fable 5. A preview is available now through Alibaba's Token Plan, Qoder, and QoderWork at 10% of standard price, and the company says it will release open weights 'soon.'
On substance, developer Shuai Bai described it as the team's first multimodal model with over a trillion parameters, handling images, videos, and documents. Alibaba claims it outperforms its own Qwen 3.7-Max, especially in coding and complex productivity tasks. The release is read as a response to momentum around Moonshot AI's Kimi K3; per Bloomberg, Moonshot reached $300 million in annual recurring revenue in June and may go public within six months.
The abundance case rests on the open-weight promise. When a model this capable is released with downloadable weights, developers, researchers, and smaller organizations can run and adapt it without paying per-query fees to a single vendor — which keeps high-end AI capability broadly available rather than gated behind one company's pricing. That is the difference between a demo and a durable public resource.
But the caveats matter here. No benchmark results are yet available, and all performance claims come from the Qwen team rather than independent testing. Most importantly, the open weights have not actually been released — only promised. Until they ship and outside evaluators can verify the claims, this is a preview, not a delivered good.
Source: The Decoder