Alibaba has made its Qwen4-Max model generally available through QwenCloud, marking a step forward from the preview stage. The release follows initial online discussion and positions the model as a major update in Chinese artificial intelligence.

The model features a 2.4-trillion-parameter backbone built on a mixture-of-experts architecture. It natively accepts images and supports a one-million-token context window, allowing it to process vast amounts of information in a single pass.
Alibaba launched Qwen4-Max. API prices fall to $1.50 input / $4.50 output per million tokens (from $2 / $6), and Alibaba says offensive-cyber training data was removed. $BABA https://t.co/ogHzlb6G9j
During demonstrations, Qwen4-Max showed ability to build playable games in Unreal Engine, design 3D automotive production lines, and execute complex reasoning tasks. Compared to its predecessor, it achieves higher efficiency with fewer tokens.
Alibaba reduced API pricing, lowering average task costs by approximately 58% on its internal Work360 benchmark.
Alibaba reduced API pricing, lowering average task costs by approximately 58% on its internal Work360 benchmark. Qwen4-Max outperforms Qwen3.8-Max by delivering higher efficiency and lower token usage during complex tasks.
Qwen just dropped Qwen4-Max and the numbers are not subtle 2.4T parameters, 71B activated, 1M-token context, API live today Work360 cost per task fell from $4.43 to $1.67, down around 62% I'd bet the "teammate not chatbot" pitch is the real story here https://t.co/CItjpZbCOP
The release includes strengthened safeguards against cybercrime and misuse of hazardous materials information. Alibaba also introduced a revision plugin for autonomous agents and plans to release smaller Qwen4 models as open source.
As the company noted upon launch, the model "can take on a wide range of creative and professional work, spanning 3D modeling, game development, precision manufacturing, finance, law, healthcare, and frontier scientific discovery."
