Alibaba has previewed its Qwen3.8 artificial intelligence model, billing it as the most powerful in its Qwen line and claiming it ranks among the world’s top-tier models after Anthropic’s Claude Fable5.

SiliconANGLE reported on July 19 that Alibaba unveiled a preview version, Qwen3.8-Max-Preview, at the World Artificial Intelligence Conference in Shanghai. The model has 2.4 trillion parameters and adopts a multimodal architecture that processes not only text but also images, video and documents. It is the first Qwen model with more than 1 trillion parameters to support multimodality.

Alibaba’s Qwen team said in a post on X, formerly Twitter, that the model is "second only to Fable5." It did not disclose benchmark scores to support the claim, and the model has not yet been verified by an independent evaluation body.

That stands out more when compared with its previous flagship model. When Alibaba released Qwen3.7-Max in May, it also published AI performance metrics and detailed results. That model scored 56.6 on the Artificial Analysis Intelligence Index. By contrast, the Qwen3.8 preview includes neither a model card nor the number of activated parameters nor benchmark data.

Alibaba also did not provide task-by-task comparison results between the new model and Qwen3.7-Max. The company said only that Qwen3.8 is "continuously evolving." It said open weights would be released "soon," but did not give a specific schedule or licensing terms.

The preview version is currently available through Alibaba’s Token Plan subscription service and its developer platforms Qoder and QoderWork. During the trial period, pricing is set at about 10 percent of the standard rate.

The timing of the unveiling also intersects with competition in China. Beijing-based Moonshot AI Technology launched Kimi K3, a model with 2.8 trillion parameters, three days earlier. The successive launches mean Chinese developers are entering competition over multi-trillion-parameter models in earnest. Until now, models of that scale have largely been seen as the domain of major U.S. AI labs.

Over the past year, Alibaba has steadily released open-weight models and has put Qwen at the center of its AI strategy. It has also expanded cloud infrastructure and dedicated chips. Based on that structure, it aims to establish itself as a foundational AI provider in China. The open-weight strategy has also helped grow the Qwen developer ecosystem.

For corporate customers, verification materials matter more than performance claims. So far, Qwen3.8 has not been evaluated on independent leaderboard rankings, and the open weights Alibaba has teased have not yet been released. While Qwen3.7-Max, the previous evaluation target, was far from the top tier on the LMArena rankings, Claude Fable5 is ranked No. 1. Ultimately, Qwen3.8’s competitiveness is expected to be determined by whether benchmarks are released and whether open weights are announced.

Qwen3.8 is launching and going open-weight soon! With a massive 2.4T parameters, this model is continuously evolving. We believe it’s one of the most powerful model available today, compatible to leading frontier AI models , second only to Fable 5. You don't have to wait to… pic.twitter.com/JS3ID73IYS