Alibaba launched Qwen4-Max on November 16, 2026, and is pitching the new model mainly on cost. The company says the model finishes tasks for far less money than its earlier flagship.
Qwen4-Max sits at the top of the new Qwen4 series. Alibaba says it uses about 50% fewer tokens than Qwen3.8-Max, and average task costs fell by more than 60% compared with that model.
Tokens are the small pieces of text an AI model reads and writes. Most AI companies charge customers based on how many tokens a job uses.
How the Pricing Works
Alibaba set the price at $1.50 per million input tokens and $4.50 per million output tokens. Input tokens are what a user sends in, and output tokens are what the model sends back.
The two savings work together. The model uses fewer tokens per job, and each token costs less, which is how Alibaba reaches the 60% figure.
Research findings on the launch put the cost per task at about $3.27. That figure is lower than the cost per task for competing models.
Alibaba is focusing the launch on efficiency and practical use rather than raw capability. The company is targeting professional productivity and coding tasks.
Size, Memory and Coding Scores
Qwen4-Max runs on a sparse mixture-of-experts design with a 2.4-trillion-parameter backbone. This setup sends each request to a small group of internal “experts” instead of using the full model every time.
That design helps keep computing costs down. It lets a very large model run without paying the full computing bill on every request.
The model is multimodal, so it can work with text, images and video. It supports a context window of up to 1 million tokens, which acts like short-term memory for long documents or large codebases.
On the DeepSWE v1.1 coding benchmark, Qwen4-Max scored 75.7. Alibaba said that score beats Claude Opus 5.5 and is competitive with Kimi K3.1 Pro.
Independent rankings place Qwen4-Max behind competitors such as Kimi on overall performance. Those same rankings list its favorable cost per task as a strength.
Alibaba is aiming the model at sectors such as healthcare, finance and law. Teams in these fields often handle large amounts of documents and watch costs closely.
Developers can use Qwen4-Max through an API on QwenCloud. The model also works with protocols set by OpenAI and Anthropic, so teams on those platforms can switch with fewer changes.
Alibaba plans to release smaller open-weight models in the Qwen4 series. Qwen4-Max itself will remain closed.
A Token Plan preview opened on November 11, 2026, before the public launch. The preview runs until November 25.