Qwen4-Max: subscription only, no per-token price, best before 25/11

In short

• Through the API, Qwen4-Max-Preview is only available with the Token Plan, Alibaba Cloud's subscription. "Pay-as-you-go access is not supported," the model page says. Preview from 11/11/2026 at 10:00 to 25/11/2026 at 24:00, UTC+8: from 11/11 at 03:00 to 25/11 at 17:00, Paris time. Then it is retired.
• Personal Edition prices: $6 a month for the Lite tier ($3 for new customers until 30/11/2026), $10 for Essential ($5), $18 for Standard, $43 for Pro. From 20,000 to 200,000 Credits a month. How many Credits one Qwen4-Max-Preview request costs: not published.
• Same house, same week: Qwen Code 1.0 Preview, Alibaba's open-source coding tool, announced on 11/11/2026 according to the Chinese press.

One till, and it charges by the month

There are two ways to pay for an AI model. Pay-as-you-go: every token read or written is billed, the bill tracks what you use, to the cent. Subscription: a monthly fee, a quota, and when the quota runs dry, the service stops.

For Qwen4-Max-Preview, Alibaba made the choice for you. Access through the Token Plan only. Pay-as-you-go is not supported. The Token Plan is the subscription of Model Studio, Alibaba Cloud's AI platform. It is billed monthly, quarterly or yearly, and counted in Credits, an in-house currency. Two editions, Personal and Team. Both run in a single region: Singapore.

Outside the API, Qwen announces early access in its own products: QwenWork, Qoder, Qwen APP and QwenStudio. Its announcement post on X, published on 11/11 at 02:01 UTC, showed 1,054,598 views the same day at 20:24 UTC. The rallying cry: "Put it to the test on a real problem". The first real problem is the bill.

$3 for how many requests?

The price table is clear. Lite: $6 a month, 20,000 Credits. Essential: $10, 45,000 Credits. Standard: $18, 80,000. Pro: $43, 200,000. For an account that has never bought a Personal subscription, Lite drops to $3 and Essential to $5, until 30/11/2026 at 23:59:59 UTC+8, which is 16:59 in Paris. An extra pack of 20,000 Credits: $15. On the Team side: $20, $75 or $180 per seat per month. Qwen4-Max-Preview is on the Personal Edition's model list, with no distinction between tiers.

That leaves the only question that matters. How many Credits does one Qwen4-Max-Preview request eat? The page's answer: it depends on the model, the number of tokens, the thinking mode and the tools called. The details are in the console, on the subscription page. The model's standard rate is written neither on the model page nor on the pricing pages. And the Personal plan is for interactive use inside coding and agent tools only: no automation scripts, no custom application backend. The only worked example in the docs is for another model, qwen3.6-plus: about 3.18 Credits for one request with 8,349 input tokens, 40,794 tokens read from cache and 573 output tokens.

Then comes the promo. During a limited-time promotion, Qwen4-Max-Preview's consumption goes as low as 20% of the standard rate: five times more usage, Alibaba says. From 22:00 to 08:00 UTC+8, another 60% off: 8% of the standard rate. That night falls between 15:00 and 01:00 in Paris. 8% of what? Of the standard rate. The one that isn't published.

14 days, 14 hours, then retirement

A preview is a trial version released before the final one. This one has its dates, to the minute: from 11/11/2026 at 10:00 to 25/11/2026 at 24:00, UTC+8. That is from 11/11 at 02:00 to 25/11 at 16:00 UTC; in Paris, from 11/11 at 03:00 to 25/11 at 17:00. Fourteen days and fourteen hours. 350 hours. During that time, no changes, the page says. After that, the model will be retired. Final version, final price, date: nothing on the page.

On 13/11 at 10:30 UTC, it has 12 days, 5 hours and 30 minutes left.

The plans, for their part, are paid by the month at the very least. The Personal Edition does not, as of today, support unsubscribing. Unused quota does not carry over to the next month. The subscription also opens the other models on the list, from qwen3.8-max to DeepSeek and GLM. The model in the announcement, for its part, has a retirement date.

What it reads, what it writes

The model page gives two figures. Context: 1M tokens, about a million. Context is everything the model reads in one go: your question, your files, the conversation history. A token is a piece of a word, the unit the model counts.

Maximum output: 393,216 tokens, shared by thinking and the final answer. Qwen4-Max-Preview thinks before it answers, and by default at the highest of the three levels on offer (low, medium, xhigh). Those thinking tokens come out of the same ceiling as the answer. Alibaba says so itself: with the OpenAI-format API, the max_tokens limit now covers both, and a limit sized for the answer alone may truncate the output. The model also speaks Anthropic's API format. And the pricing page lists the thinking mode among the things that change Credits consumption. By how much: not written.

Same week: Qwen Code 1.0 Preview

Alibaba is shipping everything at once. On 11/11/2026, according to the Chinese press, it also announced Qwen Code 1.0 Preview. Qwen Code is its coding tool that runs in the terminal. Version 1.0 Preview brings together the agent harness capabilities of Qoder, its agentic coding platform, and of Qwen Code, in an open-source release. The harness is the plumbing that connects the model, the tools and the execution. Standalone installer; Node.js 22 or later only if the installer falls back to npm. 1.0 Preview ships on a separate branch, and the stable branch stays. On first launch, the connection menu offers an Alibaba ModelStudio entry, then Token Plan. The till is already plugged in.

What we don't know yet

• Qwen4-Max-Preview's standard rate in Credits, and so the number of requests $3 pays for.
• Quarterly and yearly prices: the page mentions them, without amounts.
• What happens after 25/11: final version, price, date. Nothing on the model page.
• What the model is worth: neither the model page nor Qwen's post gives a score.

Verdict: a flagship model, a single till, no per-token price, 350 hours to live. Alibaba wrote the retirement date down to the minute. The price of a request, nowhere. I have the receipt.

Words to remember

• Preview: a trial version released before the final one; for Qwen4-Max-Preview, from 11/11/2026 to 25/11/2026, then retirement.
• Pay-as-you-go: you pay for every token used, and the bill tracks real usage; not available for this model.
• Subscription (Token Plan): an Alibaba Cloud monthly plan that grants a quota of Credits; when the quota runs out, the service stops until the next month, unless you buy an extra pack.
• Credits: the Token Plan's internal currency; each request consumes some, depending on the model, tokens, thinking and tools.
• Token: a piece of a word, the unit a language model counts and bills on.
• Context: the number of tokens a model reads in one go; about a million for Qwen4-Max-Preview.
• Maximum output: the ceiling on tokens a model writes for one request; here, thinking and answer included.
• Thinking: the reasoning a model writes for itself before answering; its tokens count toward the output.
• API: the door through which a program sends its requests to a service.
• Open source: software whose code is public, which anyone can read, modify and reuse under its licence.
• Agent harness: the software that connects a model to its tools and carries out its actions.
• UTC, UTC+8: Coordinated Universal Time, and the Beijing and Singapore time zone, 8 hours ahead; Paris is on UTC+1 in November.

Source sheet

• Alibaba Cloud, Model Studio page for Qwen4-Max-Preview (last updated 11/11/2026), read on 11/11/2026 at 08:35 UTC and again at 20:23 UTC, identical text.
• Alibaba Cloud, Token Plan pages: overview, Personal Edition and Team Edition with their prices, read on 11/11/2026 at 20:23 UTC.
• Qwen Code, documentation (Overview page), read on 11/11/2026 around 17:35 UTC.
• IT之家, article of 11/11/2026 at 12:20 (Beijing time) on Qwen Code 1.0, read on 11/11/2026 at 20:24 UTC.
• Qwen (@Alibaba_Qwen) on X, post of 11/11/2026 at 02:01 UTC, reread on 11/11/2026 at 20:24 UTC.
• Links: Qwen4-Max-Preview page at Alibaba Cloud · Token Plan prices, Personal Edition · Token Plan prices, Team Edition · Token Plan overview · Qwen Code documentation · IT之家 article on Qwen Code 1.0 · Qwen's post on Qwen4-Max

Source passages

Pages read on 11/11/2026 between 08:35 and 20:24 UTC (dated copies in the article folder).

Show the 8 sources

Ref.	Quote (original language)	URL
S1↑1↑2↑3↑4↑5↑6+5	"Last Updated:Nov 11, 2026" ; "Qwen4-Max-Preview is the next-generation flagship multimodal model in the Qwen family" ; "Qwen4-Max-Preview is currently in preview and is available through Token Plan only. Pay-as-you-go access is not supported. The preview period runs from 10:00 on November 11, 2026 to 24:00 on November 25, 2026 (UTC+8); no changes will be made during this period, and the model will be retired after the preview ends." ; "Compatible with OpenAI Chat Completions, Responses, and Anthropic Messages APIs" ; "Context length | 1M tokens" ; "Maximum output | 393,216 tokens, shared by thinking and the final answer" ; "Choose Token Plan Personal or Token Plan Team" ; "Starting with qwen4-max-preview, max_tokens is equivalent to max_completion_tokens. Both limit the combined tokens for thinking and the final answer. Thinking consumes part of the same budget, so a limit previously sized for the answer alone may truncate output." ; "Thinking: Enabled by default. Choose low, medium, or xhigh (default)."	https://www.alibabacloud.com/help/en/model-studio/qwen4-max-preview
S2↑1↑2↑3↑4↑5↑6+7	"Last Updated:Nov 11, 2026" ; "Token Plan Personal Edition is currently available only in the Singapore region." ; "Lite plan" "$6/month" "Limited-time new-user price" "$3/month" ; "Essential plan" "$10/month" "$5/month" ; "Standard plan" "$18/month" ; "Pro plan" "$43/month" ; "20,000 Credits" "45,000 Credits" "80,000 Credits" "200,000 Credits" ; "The limited-time new-user price is available until 23:59:59 on November 30, 2026 (UTC+8) and is limited to accounts that have never purchased a Token Plan Personal Edition subscription." ; "Extra Bundle: Priced at $15/bundle/month, with 20,000 Credits per bundle." ; "When the plan quota is used up and no usage bundle Credits are available, the service is paused until the next subscription month." ; "Unused quota within a subscription month does not carry over to the next cycle." ; "The Credits consumed per request are dynamically determined by model type, token usage, thinking mode, and tool calls. Refer to the subscription page usage details in the console for actual consumption." ; "qwen4-max-preview preview benefits: [...] During the limited-time promotion, Credits consumption is as low as 20% of the standard rate (5x usage); calls between 22:00 and 08:00 (UTC+8) enjoy an additional 60% off, that is, 8% of the standard rate." ; model list: "qwen4-max-preview", "qwen3.8-max", "deepseek-v4-pro", "glm-5.3"… ; "The Personal Edition does not currently support unsubscription." ; "This plan is intended for interactive use only within coding tools and agent tools such as Claude Code, Cursor, Qwen Code, Qoder and OpenClaw. It must not be used for automation scripts, custom application backends, or any non-interactive batch call scenarios."	https://www.alibabacloud.com/help/en/model-studio/token-plan-personal-overview
S3↑1↑2↑3↑4	"Token Plan Team Edition currently supports only the Singapore region." ; "Standard | USD 20/seat/month" ; "Advanced | USD 75/seat/month" ; "Premium | USD 180/seat/month" ; "Using qwen3.6-plus as an example": "Input tokens | 8,349 | 1.67" ; "Cached tokens | 40,794 | 0.82" ; "Output tokens | 573 | 0.69" ; "Approx. 3.18 Credits"	https://www.alibabacloud.com/help/en/model-studio/token-plan-team-overview
S4↑1↑2↑3↑4	"Token Plan is an AI model subscription service from Alibaba Cloud Model Studio, billed in Credits" ; "Choose a tier and subscribe with monthly, quarterly, or yearly billing." ; "Qoder | Agentic coding platform for real-world software development" ; "Qwen Code | Open-source CLI coding tool"	https://www.alibabacloud.com/help/en/model-studio/token-plan-overview
S5↑1↑2↑3↑4↑5	"Qwen Code 1.0 Preview brings together the agent harness capabilities of Qoder and Qwen Code in an open-source release." ; "Qwen’s agentic coding tool that lives in your terminal" ; "The recommended installer uses a standalone archive when one is available for your platform. If it falls back to npm, Node.js 22 or later with npm must be available on PATH." ; "For the Alibaba Cloud Token Plan ( intl ), choose Alibaba ModelStudio → Token Plan" ; "Last updated on November 11, 2026"	https://qwenlm.github.io/qwen-code-docs/en/users/overview/
S6↑1↑2↑3↑4↑5↑6	"2026/11/11 12:20:07" ; "IT之家 11 月 11 日消息，阿里今日官宣推出 Qwen Code 1.0 版本，Qoder 与 Qwen Code 融合开源。" (IT Home, 11 November: Alibaba today officially announced Qwen Code 1.0, merging Qoder and Qwen Code in open source) ; "Harness 负责衔接模型、工具与执行过程" (the harness connects the model, the tools and the execution process) ; "Qwen Code 1.0 Preview 以独立分支开放，现有稳定分支继续保留。" (Qwen Code 1.0 Preview opens as a separate branch; the existing stable branch is kept)	https://www.ithome.com/1/025/393.htm
S7↑1↑2↑3↑4↑5	"Qwen4-Max is now in preview on Alibaba’s Token Plan. You can also get early access to the new model on QwenWork, Qoder, Qwen APP, and QwenStudio." ; "Put it to the test on a real problem." ; "created_at_ms":1794362469000 (11/11/2026 02:01:09 UTC) ; "views":{"count":"1054598"} (read on 11/11/2026 at 20:24 UTC)	https://x.com/Alibaba_Qwen/status/2120335993643684315
R1↑1↑2↑3↑4↑5↑6	Our calculation. Preview: 11/11/2026 10:00 UTC+8 = 02:00 UTC = 03:00 in Paris (UTC+1) ; 25/11/2026 24:00 UTC+8 = 26/11 00:00 UTC+8 = 25/11 16:00 UTC = 17:00 in Paris ; length 14 d 14 h = 350 h. End of the new-customer price: 30/11/2026 23:59:59 UTC+8 = 15:59:59 UTC = 16:59:59 in Paris. Night discount: 22:00 to 08:00 UTC+8 = 14:00 to 00:00 UTC = 15:00 to 01:00 in Paris. From 13/11/2026 10:30 UTC to 25/11/2026 16:00 UTC: 12 d 5 h 30 min. X post: 1794362469 s = 11/11/2026 02:01:09 UTC. IT之家: 12:20:07 Beijing time = 04:20 UTC.	—