started · updated
Alibaba launches Qwen4-Max AI model with 2.4 trillion parameters
Alibaba has released Qwen4-Max, the flagship model of its Qwen 4 family, featuring a 2.4 trillion parameter architecture. Utilizing an updated Mixture-of-Experts design, the model activates 71 billion parameters per token, representing a 25 percent reduction in activated parameters compared to the previous Qwen3.8-Max.
The new model includes a one-million-token context window, native vision capabilities, and a 328-billion-parameter n-gram embedding layer. It is designed to handle complex coding, professional workflows, and agentic tasks.
Qwen4-Max became generally available on November 16, 2026. It is accessible via QwenCloud, QwenWork, and QwenStudio. Alibaba has indicated that smaller-scale, open-weight versions of the model are expected to be released in the coming months.