The newest flagship artificial intelligence model from Alibaba earned a score of 43 on the Artificial Analysis Intelligence Index, surpassing the 34 points recorded by the previous Qwen3.8-Max. The 2.4 trillion parameter model reduced the average cost per task to $3.27, a roughly 67% decrease from the $10.02 required by its predecessor. These gains in efficiency helped Qwen4-Max secure a No. 1 ranking among Chinese models on the Agent Arena leaderboard, where it placed 14th overall globally.
Despite the performance jump, the model still trails some peers, scoring three points lower on the intelligence index than Moonshot AI’s Kimi K3.1 Pro. Alibaba’s current API pricing for the mixture-of-experts model is $1.50 per million input tokens and $4.50 per million output tokens. To prevent the model from being used for hazardous purposes, the company filtered its training data to remove capabilities related to offensive cyber attacks and chemical, biological, radiological, and nuclear risks.




