NEWS / NOV.2026
Alibaba opens the Qwen4-Max API at lower rates
Alibaba has opened the Qwen4-Max API on QwenCloud and charges 25% less for standard input and output than it does for Qwen3.8-Max. On its internal Work360 benchmark, the company reports average savings of about 60% per task, partly because the new model uses fewer tokens.

Alibaba has opened the Qwen4-Max API with input and output rates 25% lower than those for Qwen3.8-Max
Alibaba opened the Qwen4-Max API on QwenCloud on November 16, 2026, with standard rates a quarter lower than those for Qwen3.8-Max. Developers can now integrate the model into their applications following a period of limited access that began on November 11, according to TechNode.
QwenCloud charges separately for what an application sends to the model and what the model produces. The platform charges $1.50 per million input tokens and $4.50 per million output tokens for Qwen4-Max. The same volumes cost $2 and $6, respectively, for Qwen3.8-Max.
The price gap widens when multiple requests share a prefix. If QwenCloud finds that prefix in its implicit cache, a million reused input tokens cost $0.075 with the new model, down from $0.25 previously. The 70% reduction applies to that portion of the request.
On Work360, its internal benchmark for workplace tasks, Alibaba also measures token usage. The company reports that in high reasoning mode, Qwen4-Max uses 35% fewer tokens than its predecessor. This lower token usage adds to the savings from lower rates. Alibaba puts the average savings at about 60% per task. The costs TechNode reports for this test are 11.16 yuan with Qwen4-Max and 26.58 yuan with Qwen3.8-Max, a reduction of about 58%.
For a team using the API, a response that requires another call incurs an additional cost. The team can calculate the cost of a usable result by comparing the two models on its own tasks and applying the same quality criteria to both. For each model, the team divides the total API bill, including retries, by the number of accepted results.