Alibaba (Qwen) · 2026-11-16 · seismic
Qwen4-Max — Alibaba's Qwen4 flagship is live on the API at $1.50 / $4.50
In short
Qwen4-Max is Alibaba's first full Qwen4 model: a new MoE with native vision and a 1M-token context. It is live on the QwenCloud API at $1.50 / $4.50 per million tokens, cheaper than Qwen3.8-Max.
Alibaba (Qwen) · 2026-11-16 · 6 sources checked

Alibaba's first full Qwen4 model goes live on the API, cheaper than the Qwen3.8-Max it replaces.
LIVE · updated 06:11 UTC
What changed vs Qwen4-Max-Preview
- Cheaper per token. $1.50 / $4.50 per million tokens, against $2 / $6 for Qwen3.8-Max; implicit cache input drops from $0.25 to $0.075.
- Fewer tokens per task. Alibaba says Qwen4-Max beats Qwen3.8-Max with up to about 50% fewer tokens, and average cost per task falls by more than 60% in its internal evaluation.
- Out of preview. Qwen4-Max-Preview was Token Plan only; qwen4-max is on the pay-as-you-go API as well as Token Plan Personal and Team.
Qwen4-Max-Preview vs Qwen4-Max
| Qwen4-Max-Preview | Qwen4-Max | |
|---|---|---|
| Released | 2026-11-11 | 2026-11-16 |
| Context window | 1M tokens (max output 393,216 tokens, shared by thinking and the answer) | 1M |
| Inputs | Text, Image, Video | Text, Image, Video |
| License | Proprietary (preview); no license or open-weight release announced. | Proprietary (hosted API) |
Live updates
- First independent benchmark: Artificial Analysis scores Qwen4-Max 43 on its Intelligence Index.
- Artificial Analysis scores Qwen4-Max 43 on its Intelligence Index (median of comparable models: 41), at $3.27 per index task and 127M output tokens to run the index. artificialanalysis.ai
- Explainer published: what it is, who gets it, how it compares.
- Alibaba's Qwen team announces Qwen4-Max: API live on QwenCloud, $1.50 / $4.50 per million tokens, and open weights of smaller Qwen4 models promised later. x.com
Key specs
| Context window | 1M tokens |
|---|---|
| Price (input / output) | $1.50 / $4.50 per 1M |
Quick facts
| Maker | Alibaba (Qwen) |
|---|---|
| Model ID | qwen4-max |
| Architecture | New MoE with native vision |
| Input | Text, images, video |
| Thinking | On by default, reasoning_effort high |
| Access | Pay-as-you-go API + Token Plan |
| Weights | Closed; smaller Qwen4 models promised open |
Pricing
| Input | $1.50 / 1M tokens |
|---|---|
| Output | $4.50 / 1M tokens |
| Implicit cache input | $0.075 / 1M tokens |
| Explicit cache read | $0.075 / 1M tokens |
What is it?
Qwen4-Max is the general release of Alibaba's next Qwen-Max flagship, five days after a Token Plan-only preview. Qwen's docs call it the first model in the Qwen4 family, combining reasoning, native vision and long-running task execution. Alibaba pitches it as an AI teammate for work, creative tasks and science, from 3D modeling and game development to finance, law and healthcare.
How does it work?
Under the hood is a new Mixture-of-Experts architecture that reads text, images and video and has a 1M-token context window. Thinking is on by default, with reasoning_effort set to high, and can be set to low or xhigh. The endpoint speaks OpenAI Chat Completions, OpenAI Responses and Anthropic Messages, and supports function calling, structured output, web search and built-in tools such as a code interpreter.
Why does it matter?
Cost is the headline for teams running agents: Alibaba says the new model beats Qwen3.8-Max while using up to about 50% fewer tokens, and that average cost per task drops by more than 60% in its internal tests. Those numbers are Alibaba's own and have not been checked by an outside evaluator yet. Because the API copies the OpenAI and Anthropic formats, existing coding agents can switch to Qwen4-Max by changing the model id.
Who is it for?
developers running coding and office agents on a budget
Frequently asked questions
- How much does Qwen4-Max cost?
- Qwen4-Max costs $1.50 per million input tokens and $4.50 per million output tokens on QwenCloud. Implicit cache hits cost $0.075 per million tokens, 70% less than on Qwen3.8-Max. Explicit cache creation costs $1.875 per million tokens for a 5-minute cache and $3 for a 1-hour cache. Token Plan Personal and Team subscribers can also use it.
- How is Qwen4-Max different from Qwen4-Max-Preview?
- Qwen4-Max-Preview opened on November 11, 2026 through Alibaba's Token Plan only, with no per-token price, and is retired after November 25. Qwen4-Max is the general release with the model id qwen4-max: it is on the pay-as-you-go API at published prices as well as on Token Plan. Qwen's docs list the preview as a separate model with its own preview benefits.
- Does Qwen4-Max beat Qwen3.8-Max?
- Alibaba says Qwen4-Max outperforms Qwen3.8-Max while using up to about 50% fewer tokens, and that average cost per task falls by more than 60% in its internal evaluation. It is also cheaper per token: $1.50 / $4.50 against $2 / $6 for Qwen3.8-Max. On the independent Artificial Analysis Intelligence Index, Qwen4-Max scores 43, above the median of 41 for comparable models.
- Is Qwen4-Max open source?
- Qwen4-Max is not open-weight; it is served only through Alibaba's API and apps. The Qwen team says open weights of smaller Qwen4 models will be released later, but it has not given a date or named the sizes. Developers who need downloadable weights today still have the open Qwen3.8 family.
- Where can I use Qwen4-Max?
- Qwen4-Max is available through the QwenCloud API with the model id qwen4-max, and through Token Plan. The Qwen team says people can also try it in Qwen Studio, QwenWork and Qoder. Qwen's docs list pay-as-you-go access in Singapore, Beijing, and global regions including Frankfurt, Hong Kong, Tokyo and US (Virginia).
Try it
Call model qwen4-max on QwenCloud, or try it in Qwen Studio, QwenWork or Qoder