Qwen4-Max launched: Alibaba's closed AI model cuts API prices 25%

Representative image: LUNEMax / Pixabay
Alibaba launched Qwen4-Max on 16 November 2026, a closed-weights AI model priced at $1.50 per million input tokens, 25% below its predecessor, the company says.
What Happened
Alibaba launched Qwen4-Max, the new flagship model in its Qwen family of artificial intelligence (AI) models, on Monday, 16 November 2026. The company says the model has 2.4 trillion parameters in total but uses only 71 billion of them for each word-piece it processes, can read up to one million tokens of text at a time, and understands images natively. Unlike several earlier Qwen models, Qwen4-Max's weights are closed: developers can use it only through Alibaba's cloud services, at $1.50 per million input tokens and $4.50 per million output tokens, 25% less than its predecessor, Qwen3.8-Max.
Qwen4-Max: key facts
- Launch date: Monday, 16 November 2026.
- Size: 2.4 trillion total parameters, with 71 billion active per token, according to Alibaba.
- Design: a mixture-of-experts (MoE) model, which switches on only part of its network for each piece of text.
- Context window: up to one million tokens; it also processes images.
- Price on QwenCloud: $1.50 per million input tokens and $4.50 per million output tokens, down from $2 and $6 for Qwen3.8-Max, launched in August.
- Access: through the QwenCloud API and Alibaba's Qwen Studio, QwenWork and Qoder apps; the weights are not released, though smaller Qwen4 models are due to get open weights later.
- Alibaba's own benchmarks: 75.7 on the DeepSWE v1.1 coding test against 74.2 for Claude Opus 5.5, but 48.5 on Terminal-Bench 4.0 against 59.6 for Opus 5.5 and 67.2 for Kimi K3.1.
Qwen4-Max: why it matters
A few terms first. A parameter is one of the numbers a model learns during training; more parameters usually mean more capacity. A token is a small chunk of text, often part of a word, and AI companies charge by the million tokens. A mixture-of-experts model is built from many smaller sub-networks, or "experts", and uses only a few of them for each token. That is how a model with 2.4 trillion parameters can run at the cost of a much smaller one.
The price cut is the headline for developers. Businesses that build chatbots, coding assistants and document tools on top of large models pay for every token, so a 25% cut in list price adds up quickly. Alibaba also claims the model needs fewer tokens to finish a task: on its own Work360 test, it says the average cost per task fell from $4.43 to $1.67. These are the company's figures from its own test; independent checks were not available at launch.
The closed weights are a shift. Several earlier Qwen models were released with open weights, which anyone can download and run on their own computers. Keeping the flagship closed means Qwen4-Max competes directly with other paid, cloud-only models, while Alibaba says open weights will come for smaller Qwen4 models.
For Indian developers and start-ups, which often choose AI models on cost, the launch adds another lower-priced option at the top end. Alibaba's international Model Studio offers free trial tokens in its Singapore region, according to reports. As with any AI service, companies handling Indian users' personal data should check where the data is processed under India's data protection rules.
Peepals calculation: Qwen4-Max uses about 3% of its parameters for each token (71 billion ÷ 2,400 billion = 0.0296). Both list prices fall by 25% ($2 to $1.50 and $6 to $4.50), and Alibaba's claimed cost per task on Work360 falls by about 62% ($4.43 to $1.67, since 1.67 ÷ 4.43 = 0.377). Our calculation from the reported figures.
| Item | Qwen3.8-Max | Qwen4-Max |
|---|---|---|
| Input price per million tokens | $2.00 | $1.50 |
| Output price per million tokens | $6.00 | $4.50 |
| Cost per task on Alibaba's Work360 test | $4.43 | $1.67 |
Impact
Short-term: Developers on Alibaba's cloud can switch to Qwen4-Max at a lower list price than its predecessor, though they will need to test it on their own work before relying on Alibaba's benchmark claims.
Long-term: Lower prices at the top end keep pressure on other AI providers' pricing, and the move to closed weights for the flagship may change how open-model users plan around Qwen.
Who is affected: Software developers and companies building AI products, including Indian start-ups; Alibaba Cloud customers; and rival AI model providers.
Key Takeaway
Alibaba's Qwen4-Max, launched on 16 November 2026, is a closed-weights flagship AI model that Alibaba prices 25% below its predecessor and says is cheaper per task, though its benchmark results are mixed and self-reported.
Questions and Answers
What is Qwen4-Max?
Qwen4-Max is Alibaba's flagship AI language model, launched on 16 November 2026, with 2.4 trillion parameters, 71 billion active per token and a one-million-token context window, according to the company.
How much does Qwen4-Max cost?
On QwenCloud it costs $1.50 per million input tokens and $4.50 per million output tokens, 25% less than Qwen3.8-Max.
Is Qwen4-Max open source?
No. Its weights are closed and it is available only through Alibaba's services, though Alibaba says smaller Qwen4 models will get open weights later.
Is Qwen4-Max better than other AI models?
Alibaba's own results are mixed: it scores slightly above Claude Opus 5.5 on one coding test but below Opus 5.5 and Kimi K3.1 on Terminal-Bench 4.0. Independent tests were not available at launch.
Disclaimer: Prepared by the Peepals newsroom from publicly available sources with AI assistance. Information is accurate to the best of our knowledge at the time of publication and may change. Images may be representative. Not professional advice. Report an error via our contact page.
By Anika Sharma, AI & Innovations EditorCovers AI & Innovation and more for Peepals Tech. All stories by Anika Sharma
Get the day's stories in your inbox
Peepals Weekly: the week decoded, free. One email a week, unsubscribe in one click.
Anika Sharma is a pen name of the Peepals Global Editorial Team, which researches, fact-checks and publishes every story. The portrait is an illustration.
Responsible for this story: the Peepals Global Editorial Team, led by editor Prashant Sonar. How we report









