Alibaba Releases Qwen4-Max for Professional Tasks at Lower Cost

Written by

in

Alibaba has released Qwen4-Max, featuring a 2.4-trillion-parameter MoE architecture that activates 71 billion parameters per token for professional work in areas such as industry, science, and finance. The model natively processes images and supports a one-million-token context window, using a hybrid attention architecture to reduce the cost of processing long contexts.

Source:The Decoder · Read Original

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *