Alibaba has released Qwen4-Max, a model with 2.4 trillion parameters, claiming that its average cost of completing professional tasks is 62% lower than that of Qwen3.8-Max. The model supports a 1-million-token context window, activates approximately 71 billion parameters per token, and uses 35% fewer tokens in High effort reasoning mode, with an average task cost of $1.67.
Source:CSO Online · Read original
Leave a Reply