Qwen4-Max: Alibaba's new flagship model, explained
By the ChatGBot Team · Published November 17, 2026
Alibaba has released Qwen4-Max, the first model of its new Qwen4 generation. It became available on November 16, 2026, with a 1 million token context window, native vision and a much lower cost per task than the model before it. Here is what Qwen4-Max is, how it scores, what it costs and how it compares with Claude, GPT and Kimi.
Every top AI in one app
GPT, Claude, DeepSeek, Qwen and GLM. One subscription.
Try ChatGBot →Qwen4-Max at a glance
| Qwen4-Max | |
|---|---|
| Made by | Alibaba (Qwen team) |
| Released | November 16, 2026 |
| Model ID | qwen4-max |
| Replaces | Qwen3.8-Max |
| Size | 2.4 trillion parameters, about 71 billion active per token |
| Context window | 1 million tokens |
| Max output | 131K tokens, 262K in thinking mode |
| Input | Text, images, video |
| Output | Text |
| API price | $1.50 input / $4.50 output per 1M tokens |
| Open weights | No |
What is Qwen4-Max?
Qwen4-Max is Alibaba's most capable Qwen model. It uses a mixture of experts design: the model has 2.4 trillion parameters in total, but only about 71 billion work on each token. That is about 25% fewer active parameters than Qwen3.8-Max, which helps keep it fast and cheap. Alibaba also uses a new attention design that compresses earlier text and then focuses on the parts that matter, which is how it handles a 1 million token context.
If you are new to the model family, our guide What is Qwen? explains the basics.
What is new
- Lower cost per task. On Alibaba's Work360 benchmark, the average cost per task fell 62%, from $4.43 to $1.67.
- Fewer tokens. Alibaba says it uses about 35% fewer tokens at high effort, and close to half on Terminal-Bench 4.0.
- Native vision. It reads images and video directly, not just text.
- Safer code. The share of code that both works and is secure on Last Secure Code Bench rose from 22% to 31.8%, according to The Decoder.
- Easy to plug in. QwenCloud offers OpenAI compatible and Anthropic compatible APIs.
Benchmarks
| Test | Qwen4-Max | Note |
|---|---|---|
| Artificial Analysis Intelligence Index | 43 | Qwen3.8-Max: 34, Kimi K3.1 Pro: 46 |
| Work360 (Alibaba) | 52.69 | Qwen3.8-Max: 47.74 |
| Terminal-Bench 4.0 | 48% | Kimi K3.1 Pro: 67% |
| HLE-Diamond | 45% | Up 21 points |
| Agents' Last Exam v1.1 | 46.7 | Best in Alibaba's comparison group |
| Code Arena WebDev | 6th place (1,750) | xHigh setting |
Most of these numbers come from Alibaba and were reported by The Decoder and TechEdt. Independent results may differ.
Qwen4-Max vs Claude, GPT and Kimi
Alibaba compared Qwen4-Max with Claude Opus 5.5, GPT-6 models and Kimi K3.1 Pro. The short version: it is much cheaper than the top US models, but it does not beat them on hard reasoning. It trails Opus 5.5 and GPT on Humanity's Last Exam, lands ahead of Opus 5.5 on HealthBench Professional, and Kimi K3.1 Pro stays ahead on the Intelligence Index and Terminal-Bench. Artificial Analysis puts its cost at $3.27 per index task, versus $10.02 for Qwen3.8-Max and $3.68 for Kimi K3.1 Pro.
Price and how to get access
- QwenCloud API: $1.50 per million input tokens and $4.50 per million output tokens. Cached input starts at $0.075 per million.
- Alibaba apps: QwenWork, Qoder and QwenStudio.
- Open weights: not for Max. Alibaba says smaller Qwen4 models will be released openly later.
Should you use it?
If you build with large amounts of text, documents or images and care about cost, Qwen4-Max is worth testing. For the hardest reasoning or coding tasks, Claude and GPT still lead on most tests. Trying several models on your own task is the best way to choose. See our how to use Qwen guide, and for Google's fast new model from the same week, read about Gemini 4 Hydrogen.
Try Qwen and other top models in one place
Ready to try it yourself?
Switch between top AI models in one chat.
Try ChatGBot →Frequently asked questions
What is Qwen4-Max?
Qwen4-Max is Alibaba's new flagship Qwen model, released on November 16, 2026. It is a 2.4 trillion parameter mixture of experts model with a 1 million token context window and native image and video input.
How much does Qwen4-Max cost?
On QwenCloud it costs $1.50 per million input tokens and $4.50 per million output tokens. Cached input starts at $0.075 per million tokens.
Is Qwen4-Max open source?
No. Qwen4-Max is closed. Alibaba says smaller Qwen4 models will get open weights later, but has not given a date.
Is Qwen4-Max better than Qwen3.8-Max?
According to Alibaba, yes. It scores higher on most tests, uses fewer tokens and cuts the average cost per task on its Work360 benchmark by 62%.
How does Qwen4-Max compare to Claude and GPT?
It is cheaper, but it still trails the top US models on several tests, including Humanity's Last Exam. On the Artificial Analysis Intelligence Index it scores 43.
Can I use Qwen in ChatGBot?
Yes. ChatGBot includes Qwen models next to GPT, Claude, DeepSeek and GLM, so you can compare them in one app.