Alibaba introduced Qwen3.8-Max-Preview at the World Artificial Intelligence Conference (WAIC) in Shanghai, positioning it as the most capable model in its Qwen lineup.
Alibaba said the model contains 2.4 trillion parameters. It is the first Qwen model with more than one trillion parameters to support multimodal tasks, including text, images, video, and document understanding.
In a post on X, Alibaba's Qwen team described the model as "second only to Fable 5," referring to Anthropic's flagship AI model. The preview is available through Alibaba's Token Plan, Qoder, and QoderWork platforms at 10% of the standard price during the trial period. Alibaba also said it plans to release the model's open weights "soon," though it has not provided a timeline.
Bold performance claim comes without benchmarks
Alibaba did not release benchmark scores, a model card, activated parameter counts, or detailed technical documentation alongside Qwen3.8-Max-Preview, making its performance claims difficult to independently verify.
Without independent testing, Alibaba's claim that Qwen3.8 trails only Anthropic's Fable 5 cannot yet be independently verified. Public evaluation platforms such as LMArena and Artificial Analysis have not yet assessed the new model, leaving developers without third-party data to compare its capabilities against competing systems.
Alibaba has described Qwen3.8 as "continuously evolving," signaling that the preview is still under active development.
China's AI heavyweights keep raising the stakes
The timing of the launch is significant. Alibaba unveiled Qwen3.8 just days after Chinese startup Moonshot AI introduced its 2.8-trillion-parameter Kimi K3, another large-scale model aimed at competing with leading AI systems from Anthropic and OpenAI.
The back-to-back announcements highlight how Chinese AI companies are rapidly moving into the multi-trillion-parameter category, an area previously dominated by major U.S. AI labs. Investors welcomed Alibaba's latest AI push, with the company's Hong Kong-listed shares climbing as much as 5.4% following the announcement, according to Bloomberg.
What it means for developers and businesses
For developers, Qwen3.8 offers an early opportunity to test Alibaba's newest flagship through its cloud platforms before an open-weight release arrives.
Businesses, however, may wait for more evidence before making deployment decisions. Performance claims without benchmark data make it difficult to judge how the model compares in real-world coding, reasoning, or enterprise workloads. Until Alibaba publishes technical documentation or independent researchers evaluate the model, organizations will have limited visibility into expected performance, efficiency, and operating costs.
The bigger picture
Alibaba's latest release shows that competition among Chinese AI companies is accelerating beyond simply building larger models. Increasingly, vendors are using preview launches, developer platforms, and promised open-weight releases to attract users before broader public access.
At the same time, the launch underscores a growing challenge across the AI industry: marketing claims are arriving faster than independent validation. Qwen3.8 could ultimately become one of the industry's strongest models. Until Alibaba publishes benchmark results and technical documentation—or independent evaluators validate its performance—developers will need to treat its ambitious positioning as a claim rather than a confirmed ranking.
Other News: China has launched a new global AI governance organization backed by 29 countries, signaling a growing push to shape international AI rules outside Western-led frameworks.