Alibaba has launched Qwen4-Max, a new flagship model in its Qwen family, with 2.4 trillion parameters, native vision capabilities and a one-million-token context window. The model activates 71 billion parameters per token and is available through QwenCloud’s API and Token Plan, as well as QwenWork and QwenStudio, according to Alibaba’s November 16 announcement.

Alibaba says it trained the model to retain cybersecurity knowledge that helps identify and fix software vulnerabilities while filtering out material intended to improve offensive hacking. The company also added dedicated safety datasets and post-training measures designed to reduce risks involving chemical, biological, radiological and nuclear threats.

The company reported that the share of generated code passing security checks at the function, file and repository levels rose from 8% with Qwen3.8-Max to 14.7% with Qwen4-Max. It also said internal testing showed average task costs fell by around 60%. Reuters reported the figures but said the findings had not been independently verified. Alibaba’s announcement says Qwen4-Max used 35% fewer tokens than its predecessor in the proprietary Work360 benchmark’s high-effort reasoning mode.

Qwen has already been selected to support some Apple Intelligence experiences in China, while Alibaba has been expanding its cloud infrastructure to support AI workloads. Reuters’ report on the launch also details the company’s cybersecurity and CBRN safeguards.

Alibaba says it plans to release smaller open-weight models from the Qwen4 family to developers, but has not provided a release timetable.


Buy ExpressVPN with PayPal or Credit Card
READ
Coles Probes Cyber Incident, No Evidence of Data Compromise
Advertisement