Story perspectives

Alibaba Unveils Qwen3.8-Max as US Probes Chinese AI

8/4/2026

Summary

Alibaba Unveils

Alibaba Group Holding unveiled Qwen3.8-Max, its largest AI model, on 3 August 2026.
It uses a mixture-of-experts design that activates roughly 95 billion parameters per request.
Qwen3.8-Max supports text, image and video inputs and can handle up to 1 million tokens per context.
Alibaba said the model ranked first among Chinese text models, second worldwide for visual-analysis on Arena AI, and its Hong Kong shares rose 7.7% to 126 HKD while New York shares gained 4.5%.

Summary

Bessent Probes AI

Treasury Secretary Scott Bessent announced an investigation into Chinese AI models distilled from U.S. models.
U.S. authorities accuse Chinese firms of smuggling Nvidia H200 AI chips despite export controls.
Beijing promotes open-weight releases like Qwen3.8-Max to expand AI governance influence, prompting reports of a crackdown on open tools.
AI companies rally to preserve open-weight model access, while OpenAI and Anthropic face scrutiny after reports of cyberattacks by AI agents.

History

Historical Context

After the August 2018 NDAA barred Huawei and ZTE from U.S. government use, the Commerce Department placed Huawei and 70 affiliates on the Entity List on 15 May 2019, cutting U.S. chip supplies and spurring China’s chip market to capture 35 percent of global demand by 2021. Today, Alibaba’s Qwen 3.8-Max launch and Scott Bessent’s probe echo that split as regulators target AI chip transfers.

1 of 3

Alibaba Unveils

Alibaba Group Holding unveiled Qwen3.8-Max, its largest AI model, on 3 August 2026.
It uses a mixture-of-experts design that activates roughly 95 billion parameters per request.
Qwen3.8-Max supports text, image and video inputs and can handle up to 1 million tokens per context.
Alibaba said the model ranked first among Chinese text models, second worldwide for visual-analysis on Arena AI, and its Hong Kong shares rose 7.7% to 126 HKD while New York shares gained 4.5%.

1 / 3