Alibaba Qwen4-Max cuts task cost 60 percent
Alibaba released Qwen4-Max, its new top AI model, on the 16th of November 2026. Alibaba said it cuts the cost of a typical work task by about 60 percent.

Alibaba releases Qwen4-MaxAlibaba released Qwen4-Max, its new top AI model, on the 16th of November 2026. Alibaba said it cuts the cost of a typical work task by about 60 percent.
The model reads text, images and video and answers in text. It can take in about one million tokens at once, which is a long book's worth of words.
A giant model that wakes up a small part
Qwen4-Max has 2.4 trillion parameters, the settings a model learns in training, the Qwen team said. Only 71 billion of them run for each word piece, which keeps it cheaper to use.
Alibaba did not release the model for download. It said smaller Qwen4 models will come out free to download later, with no dates given.
How it compares
On Work360, Alibaba's own work test, Qwen4-Max scored 52.69 against 47.74 for the older Qwen3.8-Max. It also uses about 24 to 49 percent fewer tokens on some tests, Alibaba said.
Alibaba said it beat Anthropic's Claude Fable 5.1 and Moonshot's Kimi K3.1 Pro on coding tests, The Straits Times reported. Those are company claims, and outside testers have not yet checked them.
Price
The model costs 1.50 dollars per million input tokens and 4.50 dollars per million output tokens, the Sulat news site reported from Alibaba's price page.
Below we break down the cost per task
In Alibaba's test, an average task fell from 4.43 dollars to 1.67 dollars, Sulat reported. That saves 2.76 dollars per task, a cut of about 62 percent, by our arithmetic.
IT Home reported the same test in Chinese yuan, falling from 26.58 to 11.16, a cut of 58 percent.


Key facts
Alibaba released Qwen4-Max on the 16th of November 2026.
It has 2.4 trillion parameters and reads about one million tokens.
Input costs 1.50 dollars per million tokens.
Alibaba says a typical task costs about 60 percent less.