Qwen4-Max is Alibaba’s newly released flagship model for multimodal and agent work. The practical change is production access; the right migration depends on the endpoint, region and features an application actually uses.
Read the dated primary source ↓What Alibaba released
Alibaba announced Qwen4-Max on November 16, 2026. The company describes a mixture-of-experts model with 2.4 trillion total parameters and 71 billion active per token, native vision and a one-million-token context window. It is available through QwenCloud’s API and Token Plan, with access in QwenWork and QwenStudio.
Alibaba reports that, on its proprietary Work360 benchmark, the model uses 35% fewer tokens at High effort and reduces average task cost by roughly 60% compared with Qwen3.8-Max. These are company-reported results on that evaluation, not a measured saving for Promptwire or a forecast for every application.
The announcement says smaller open-weight Qwen4 models are coming. That promise does not establish downloadable weights for the flagship released here.
Public record: Alibaba Cloud launch announcement — November 16, 2026
The production model and preview have different identifiers
The preview documentation now recommends the production version. The preview uses qwen4-max-preview and remains a Token Plan-only offering, without pay-as-you-go access. Its published window ends at 24:00 on November 25, 2026 (UTC+8), after which that preview model is due to retire.
That deadline belongs to the preview. Treat the production identifier qwen4-max as a separate configuration choice, and check the production documentation rather than carrying over every preview setting.
The November 16 Token Plan changelog confirms support for the production model in both Personal and Team editions.
Public record: Alibaba Cloud — Qwen4-Max-Preview access and retirement window · Alibaba Cloud — Token Plan updates, November 16, 2026
Regional API prices are not identical
The production page lists the following US-dollar rates per million tokens, checked November 17. Its output billing includes reasoning and answer tokens. The published table applies the listed rates across requests up to the one-million-token range.
| Deployment | Input / million tokens | Output / million tokens |
|---|---|---|
| Singapore / International | $1.50 | $4.50 |
| US, Germany, Japan and Hong Kong / Global | $1.413 | $4.24 |
| Beijing / Chinese mainland | $1.413 | $4.24 |
These are the input and output list rates shown on the model page. Caching, account eligibility, quotas and other service charges need their own check.
Public record: Alibaba Cloud — Qwen4-Max production specifications and pricing
Check the feature matrix before migrating
The current production documentation lists text, image and video inputs with text output, tool calling, web search and structured outputs. It supports OpenAI Chat Completions, Responses and Anthropic Messages interfaces. Batch inference and fine-tuning are marked unsupported.
Compatibility is a starting point for a trial. An application still needs to verify the exact tool schema, streaming events and failure handling it relies on. A request that returns successfully may still produce a result that the surrounding workflow cannot use.
Public record: Alibaba Cloud — Qwen4-Max production specifications and pricing
Promptwire’s take: compare completed work
A useful first comparison would use three tasks already familiar to the team: repair a small software defect, extract an agreed set of fields from a document, and answer a question that depends on a chart. Keep the inputs and completion criteria fixed, and include examples the team has not used to tune its prompts.
Record whether the work is accepted, how many retries it needs, the total token charge and the time spent correcting it. For tool workflows, include a failed tool call and inspect whether the application recovers clearly. Run the comparison against the endpoint and region you intend to use.
The release offers a new production option, but a large context window and favourable launch figures cannot decide that choice by themselves. The useful result is a repeatable improvement on the work the team actually ships. Promptwire has not independently benchmarked Qwen4-Max.
Questions readers ask
Does the preview retirement date apply to Qwen4-Max production?
The November 25 deadline is stated on the qwen4-max-preview page. It should not be treated as a retirement announcement for the separate production model.
Follow the evidence.
Primary sources checked November 17, 2026. Availability, prices and benchmark results are attributed to Alibaba. Promptwire has not independently tested the model. Migration suggestions and evaluation steps are editorial analysis. Cover: AI-assisted conceptual illustration, not a product screenshot.
- Alibaba Cloud launch announcement — November 16, 2026
- Alibaba Cloud — Qwen4-Max production specifications and pricing
- Alibaba Cloud — Qwen4-Max-Preview access and retirement window
- Alibaba Cloud — Token Plan updates, November 16, 2026
Public sources support the factual background. Analysis and illustrative exercises are identified in the text. How we assess AI · Editorial standards
Spot a factual error or an outdated detail?



