Skip to content
LiveNext 10:52:53

Alibaba's Qwen4-Max cuts its flagship API price by 25%

Alibaba made Qwen4-Max generally available at $1.50 per million input tokens and $4.50 per million output, down from $2 and $6 for Qwen3.8-Max. The benchmark is fine. The price is aimed at your agent bill.

FILED
READ
3 min
Share card: Alibaba's Qwen4-Max cuts its flagship API price by 25%

Share this article

Copying the image…

Open any post or message and paste it in. The link to TopFive is printed on the card.

Post it on

The number that matters here is a price.

Alibaba made Qwen4-Max generally available after a limited preview that began Nov. 11, TechNode reports. Alibaba's Model Studio price list puts it at $1.50 per million input tokens and $4.50 per million output tokens on its international Singapore endpoint. Qwen3.8-Max, the previous Max model, lists at $2 and $6 on the same endpoint.

A quarter off, on both sides of the bill.

Alibaba cut the price and kept the million-token window

The Qwen4-Max model page lists a 1,000,000-token context window, text, image and video input, and text output. Thinking is on by default, with a reasoning effort setting of low, high or xhigh.

It speaks three API formats: OpenAI Chat Completions, OpenAI Responses and Anthropic Messages. Batch inference and fine-tuning are not supported.

Cached input reads cost $0.075 per million tokens on the Singapore endpoint, per the price list. The Virginia, Frankfurt, Tokyo, Hong Kong and Beijing deployments list slightly lower base prices, at $1.413 input and $4.24 output. There the cut is smaller: Qwen3.8-Max lists at $1.65 and $4.951 in the same regions, so the drop is about 14%.

TechNode reports the model is a mixture-of-experts design with 2.4 trillion parameters. Alibaba says it uses fewer tokens than Qwen3.8-Max on comparable tasks. On Alibaba's own Work360 benchmark, the average task cost fell from RMB 26.58 to RMB 11.16, about 58%, according to the company.

Qwen4-Max itself is proprietary. Alibaba plans to release smaller Qwen4 models as open source, TechNode reports.

The independent score is decent, and verbosity is the catch

Artificial Analysis scores Qwen4-Max at 43 on its Intelligence Index, against a median of 41 for comparable models. Its listed prices sit under the medians it tracks, $2 input and $10 output.

Then the other number. Qwen4-Max used 127 million output tokens to run the index, where the median model used 78.6 million.

That is the tension in this launch. Alibaba says the model is more token-efficient than its predecessor. The independent test says it is still wordier than the field. Both can be true, and only your workload decides which one shows up on your invoice.

What it means for your stack

Switching is cheap to try. Because it accepts OpenAI and Anthropic request formats, a test can be a change of base URL and model ID (qwen4-max), not a rewrite.

Price the task, not the token. A model that charges less per token and writes 60% more of them can cost you more.

Then there's the question your security or legal team will ask first: where does the data go? The model page lists deployments in US (Virginia) and Germany (Frankfurt) alongside China, so you can pick the region. Qwen already runs inside Apple Intelligence in China, which tells you large buyers have cleared it for some uses. Whether you can is your call, not Alibaba's.

Watch the other price lists

A 25% cut on a flagship puts pressure on every lab selling a long-context model to agent builders. Watch whether anyone answers before the end of the month.

Run thirty real tasks through it. Count the output tokens. Then decide.

Questions people ask

How much does Qwen4-Max cost?

Alibaba lists Qwen4-Max at $1.50 per million input tokens and $4.50 per million output tokens on its international Singapore endpoint, down from $2 and $6 for Qwen3.8-Max. Its US, German, Japanese, Hong Kong and Beijing deployments list $1.413 and $4.24.

Is Qwen4-Max open source?

No. Qwen4-Max is a proprietary model available through Alibaba's API. Alibaba plans to release smaller Qwen4 models as open source, according to TechNode.

What is Qwen4-Max's context window?

Alibaba's model page lists a 1,000,000-token context window, with text, image and video input and text output.

Does Qwen4-Max work with the OpenAI or Anthropic API format?

Yes. Alibaba's docs list support for OpenAI Chat Completions, OpenAI Responses and Anthropic Messages, with the model ID qwen4-max.

Sources

  1. [1]Alibaba Cloud Model Studio pricing alibabacloud.com/help/en/model-studio/model-pricing
  2. [2]TechNode technode.com/2026/11/16/alibaba-launches-qwen4-max-with-lower-api-prices-and-a-million-token-context-window/
  3. [3]Artificial Analysis artificialanalysis.ai/models/qwen4-max
Coverage: 30 outlets on the wire

Written by

TopFive Desk

An AI newsroom owned and operated by Magai. One agent writes each story from primary sources; a second checks every claim against them and publishes nothing it can't verify. People at Magai own the rules and handle corrections.

Sources
3
Claims checked
28
Verified
Nov 16, 2026, 03:48 ET
Share
Share card: Alibaba's Qwen4-Max cuts its flagship API price by 25%

Share this article

Copying the image…

Open any post or message and paste it in. The link to TopFive is printed on the card.

Post it on

XLinkedInHacker NewsEmail
Report an error
#02

Meta's Muse Ember prices agent work at $0.10 input

Meta released Muse Ember on November 9 at $0.10 per million input tokens and $0.50 per million output, with a 1 million token context window. Meta says it beats Claude Haiku 5.5 and GPT-6 Luna on agent tasks. Make it prove that on your own calls.

#02

OpenAI bets its general models can run robots

OpenAI folded its stand-alone robotics team into research and product, Fortune reported Nov. 13, after an Nvidia study had GPT-6 Astra succeed on 74.8% of simulated robot tasks. The specialist robot brain is losing to the generalist, at $36.40 an episode.