How to use Qwen4-Max right now: subscribe to Alibaba’s Token Plan and call the model ID qwen4-max-preview from Claude Code, Qwen Code or any OpenAI- or Anthropic-compatible tool. If you don’t write code, Alibaba’s Qwen team says early access is also open in QwenWork, Qoder, the Qwen App and QwenStudio. There’s no pay-as-you-go API price yet.
That’s the short version. Qwen4-Max went into preview on November 11, 2026, and the announcement on X named five places to try it. This guide walks through each one: who it suits, what you need and what limits apply. Then it covers the developer path, where your coding agent runs on Qwen4-Max, builds a small API and tests it with Apidog CLI.
If you want the background on the model itself first, read what Qwen4-Max is. For a deep look at the subscription, see the Qwen4-Max Token Plan guide.
Every way to access Qwen4-Max: quick comparison
Qwen4-Max is a preview model, so access runs through Alibaba’s own products, not a public per-token API. Here’s how the five routes compare as of November 11, 2026.
| Route | Best for | What you need | Cost to start |
|---|---|---|---|
| Token Plan | Developers using coding agents | QwenCloud account, a plan, a Token Plan API key | $3 first month on Lite (event price; list $8) |
| Qoder | Developers who want an agentic IDE | Qoder account and app | Credit-based; check the selector |
| QwenWork | Knowledge work and file tasks | QwenWork account (web or desktop) | Free plan available |
| Qwen App | Quick tests on your phone | Qwen App and account | Not published for Qwen4-Max |
| QwenStudio | Browser chat and prototyping | Qwen account | Not published for Qwen4-Max |
Two things haven’t been published yet, so don’t plan around them: official benchmarks and a pay-as-you-go API model ID with a price. Third-party posts quote a 1M-token context window, but Alibaba hasn’t stated a context length for Qwen4-Max on its own pages. Treat that number as unconfirmed.
Route 1: Qwen4-Max on the Token Plan (the developer route)
The Token Plan is QwenCloud’s subscription for AI coding tools. You pay monthly, get a pool of Credits, and spend them on any model the plan lists. On the Token Plan pricing page, qwen4-max-preview sits second in the model list with a “Limited-Time 80% Off” tag and three capabilities: text generation, reasoning and visual understanding.
Token Plan prices (checked November 11, 2026)
The page shows these Individual monthly prices in US dollars. The lower number is the event price for new customers, and the page says the final price is whatever checkout shows after you log in.
| Plan | Event price | List price | Credits per month | Concurrent agents |
|---|---|---|---|---|
| Lite | $3.00 | $8.00 | 20,000 | 1-2 |
| Essential | $5.00 | $16.00 | 45,000 | 2-3 |
| Standard | $18.00 | $25.00 | 80,000 | 3-4 |
| Pro | $43.00 | $51.00 | 200,000 | 6-8 |
The page doesn’t say how many Credits one Qwen4-Max request burns, so you’ll learn your real usage from the plan’s usage page after a few sessions.
Rules to know before you buy
The Token Plan Individual FAQ sets a few limits worth reading twice:
- Interactive use only. The plan is for “personal interactive development through designated tools.” Using the key for production automation, batch scripts or background scheduled tasks isn’t allowed.
- No sharing. Sharing accounts or API keys is prohibited.
- No refunds. Token Plan Individual is non-refundable.
- One region. The overview says Token Plan can be bought and used only in the Singapore region, and its keys start with
sb-sp-.
That first rule matters. Your coding agent can use Qwen4-Max while you build. Your shipped app’s backend shouldn’t call the Token Plan key.
Connect Claude Code to Qwen4-Max
QwenCloud’s Claude Code setup page maps Claude Code’s model slots to Token Plan models, with Qwen4-Max in the Opus slot. Add this to ~/.claude/settings.json:
{
"env": {
"ANTHROPIC_AUTH_TOKEN": "YOUR_TOKEN_PLAN_API_KEY",
"ANTHROPIC_BASE_URL": "https://token-plan.maas.qwencloudapi.com/apps/anthropic",
"ANTHROPIC_MODEL": "auto",
"ANTHROPIC_DEFAULT_HAIKU_MODEL": "qwen3.6-flash",
"ANTHROPIC_DEFAULT_SONNET_MODEL": "qwen3.8-flash",
"ANTHROPIC_DEFAULT_OPUS_MODEL": "qwen4-max-preview",
"CLAUDE_CODE_SUBAGENT_MODEL": "auto",
"CLAUDE_CODE_MAX_CONTEXT_TOKENS": "983616"
}
}
Open a new terminal and run claude -p "Hello" to check the connection. To force Qwen4-Max for a whole session instead of auto, set ANTHROPIC_MODEL to qwen4-max-preview. If long answers get cut off, the same page says to set CLAUDE_CODE_MAX_OUTPUT_TOKENS=64000.
Connect Qwen Code or any OpenAI-compatible tool
The Qwen Code guide uses the OpenAI-compatible endpoint instead:
Base URL: https://token-plan.maas.qwencloudapi.com/compatible-mode/v1
Model: qwen4-max-preview
Key: QWENCLOUD_TOKEN_PLAN_API_KEY (from your API Keys page)
The guide notes that Qwen4-Max’s default output limit is 32,768 tokens. For longer output, set generationConfig.samplingParams.max_tokens in that model’s entry. The pricing page also lists Cline, Cursor, OpenCode, Codex, Kilo CLI and OpenClaw as supported tools.
Route 2: Qwen4-Max in Qoder
Qoder is Alibaba’s agentic coding IDE, and the announcement lists it as an early-access surface. Qoder bills in Credits, and its model selector docs say the selector shows each model’s Credit rate before you send a prompt.
As of November 11, those docs list Qwen3.8-Max but not Qwen4-Max yet. Update Qoder to the latest version and look for Qwen4-Max in the model selector. Check the rate it shows there, because Qoder ran a steep promo rate on the last Qwen Max preview and the Qwen4-Max rate isn’t documented yet.
Pick Qoder if you want an IDE built around agents and don’t want to wire up a base URL yourself.
Route 3: Qwen4-Max in QwenWork
QwenWork is an agent workspace for knowledge work. You describe an outcome, and it plans the steps, uses your files and returns a finished file. It runs in the browser or as a desktop app for Mac and Windows.
QwenWork has a $0 Free plan with 500 sign-up bonus credits and 200 daily login credits, both marked limited time. Paid plans start at $20 a month for Plus. The QwenWork site doesn’t describe a model picker, so check inside the app for Qwen4-Max before you count on it.
Choose QwenWork for reports, briefs and file tasks, not for writing code.
Route 4 and 5: the Qwen App and QwenStudio
The Qwen App (mobile) and QwenStudio (Qwen’s chat workspace) are the zero-setup routes. Sign in, open the model menu and pick Qwen4-Max if it’s listed for your account. Neither has published a Qwen4-Max quota, so expect early-access limits to change during the preview.
These two suit a first impression: ask hard reasoning questions, paste a screenshot to test visual understanding, or compare its answers with Qwen 3.8. For a full side-by-side, see Qwen4-Max vs Qwen 3.8.
Which Qwen4-Max route should you pick?
Match the route to the job:
- You write code with an agent: Token Plan. It’s the only route with a documented model ID, base URL and price.
- You want an IDE that handles setup: Qoder.
- You want finished documents, not code: QwenWork.
- You want a quick look today: QwenStudio or the Qwen App.
If you build production apps, wait for a pay-as-you-go API listing. The Token Plan’s terms rule out backend use, and preview models can change without notice. Our Qwen 3.8 API guide covers the current production option.
Build and test an API with Qwen4-Max and Apidog CLI
Here’s the developer path in practice. Your coding agent runs on Qwen4-Max through the Token Plan, writes a small REST API, then uses Apidog CLI to test it. You write one prompt. The agent does the typing.
Set up Apidog CLI
Install the CLI, log in with your personal access token and add the Apidog skill to your agent:
npm install -g apidog-cli@latest
apidog login --with-token <YOUR_TOKEN>
npx -y skills add https://apidog.com # pick Claude Code when prompted
A shortcut: tell your agent Read https://apidog.com/apidog-cli-installation-guide.md and set up Apidog CLI. That guide is written for agents. The installation guide has the human version.
Give your agent one prompt
With Claude Code pointed at qwen4-max-preview, send this:
Build a small Express API with POST /todos and GET /todos/{id}, plus an
openapi.json for it. Start it on port 3000. Then import the spec into Apidog
project <projectId>, create a test case for POST /todos that asserts a 201
status and a returned id, run it against environment <envId>, and summarize
any failures with the fix you'd make.
The apidog commands the agent runs
After it writes the code and spec, the agent imports the spec, reads the schema for a test case, validates its payload and runs it:
apidog import --project <projectId> --format openapi --file ./openapi.json
apidog cli-schema get test-case-create
apidog cli-schema validate test-case-create --file case.json
apidog test-case create --project <projectId> --file case.json
apidog run --test-case <caseId> --project <projectId> -e <envId> -r cli,json
The cli reporter prints a pass or fail summary in the terminal, and json writes a report to the apidog-reports folder, which the agent reads to explain failures. Add --upload-report if you want to view the run in the Apidog app.
This is also a fair way to judge a preview model. Without published benchmarks, a passing test run on your own API tells you more than a leaderboard would. Run the same prompt with qwen3.8-flash in the Sonnet slot and compare how many fix cycles each model needs.
The test itself doesn’t need Qwen4-Max, so you can move apidog run ... -r junit into CI without touching your Token Plan key. The Apidog CLI CI/CD guide shows the setup, and Apidog CLI in Claude Code covers more agent prompts. Apidog’s Free plan covers up to 4 users.
FAQ
Is there a Qwen4-Max API?
Not a pay-as-you-go one yet. As of November 11, 2026, the only documented API access is the Token Plan, using model ID qwen4-max-preview and the Token Plan base URL. A public per-token model ID and price haven’t been published.
How much does Qwen4-Max cost?
On the Token Plan, Individual plans run from $3 (Lite event price, list $8) to $43 (Pro event price, list $51) a month, checked November 11, 2026. The page doesn’t state Credits per Qwen4-Max request.
Can I use Qwen4-Max for free?
QwenWork has a $0 plan with limited-time bonus credits. No route documents free Qwen4-Max access yet, so check the model menu in QwenWork, the Qwen App or QwenStudio before you count on it.
What is the Qwen4-Max context window?
Alibaba hasn’t published one. Third-party reports say 1M tokens, and QwenCloud’s Claude Code setup sets a client context limit of 983,616 tokens, but neither is an official model spec.
Can I use my Token Plan key in production?
No. The FAQ bans production automation, batch scripts and background scheduled tasks. Use it for interactive coding, and keep your shipped backend on a pay-as-you-go API.
Which Qwen model should I use until Qwen4-Max is generally available?
For production APIs, use a released model with published pricing. Our guide to the best Qwen models and Qwen 3.8 for coding compare the current options.
