{
  "data": {
    "author": "Vlad Cealicu",
    "faq": null,
    "kicker": "Blog · Opinion · 11 November 2026",
    "lede": "Alibaba's Qwen team said this morning that Qwen4 is coming and that Qwen4-Max is in preview. I wanted it listed the same night. It isn't, because the only place Qwen4-Max is written down is a post on X.",
    "readingMinutes": 4,
    "words": 911
  },
  "kind": "anchor.page",
  "links": {
    "api": "https://www.anchorterminal.com/api/v1/index.json",
    "html": "https://www.anchorterminal.com/blog/qwen4-max-preview-docs-still-say-3-8",
    "json": "https://www.anchorterminal.com/blog/qwen4-max-preview-docs-still-say-3-8.json",
    "llms": "https://www.anchorterminal.com/llms.txt",
    "markdown": "https://www.anchorterminal.com/blog/qwen4-max-preview-docs-still-say-3-8.md",
    "mcp": "https://www.anchorterminal.com/mcp",
    "slim": "https://www.anchorterminal.com/blog/qwen4-max-preview-docs-still-say-3-8.min.md"
  },
  "markdown": "By Vlad Cealicu, 11 November 2026. 911 words, about 4 min to read.\n\nAt one minute past two this morning, Wednesday 11 November, Alibaba's Qwen team posted that Qwen4 is coming and that Qwen4-Max is in preview on its Token Plan, with early access in QwenWork, Qoder, the Qwen app and QwenStudio [[1]](#ref-1). I was still working, and forty minutes later I asked for it to be listed on Anchor Terminal. It isn't listed. We list what an agent can call today, from what the company's own pages say, and at three in the morning none of Alibaba's pages that I could read said Qwen4 anywhere.\n\n## What Alibaba said\n\nThe post promises \"Deeper knowledge. Richer world understanding. Stronger coding and agentic capabilities\", in a model \"built for real-world tasks\", from complex coding and knowledge work to front-end builds [[1]](#ref-1). It has no number in it. There's no context size, no benchmark, no price per token and no model id, and it ends by asking people to put the model to the test on a real problem and show what they build [[1]](#ref-1).\n\nFifty days ago, on 22 September, Alibaba said at its conference in Hangzhou that \"its next-generation model, Qwen 4, is currently in training\" [[2]](#ref-2). Going from in training to in people's hands in fifty days is fast, and I think that's the news here, more than any adjective in the post.\n\n## What I could check\n\nToken Plan is a subscription on Qwen Cloud for coding and agent tools. Lite is $6 a month with 20,000 credits, Essential $10 with 45,000, Standard $18 with 80,000 and Pro $43 with 200,000, and a plan runs one or two agents at a time on Lite and six to eight on Pro [[3]](#ref-3). A new customer pays $3 for Lite until 30 November [[3]](#ref-3). It works in Qwen Code, Claude Code, Cursor, Codex and a few others, over OpenAI's and Anthropic's protocols [[4]](#ref-4).\n\nSo the price of trying Qwen4-Max is known, and it's low. What it costs per token isn't, because credits are worked out from the model, the tokens, thinking mode and tool calls, and the docs don't have Qwen4-Max in them [[3]](#ref-3). The list of models a Token Plan covers ends at `qwen3.8-max` [[3]](#ref-3). The newest entry in Qwen Cloud's model changelog is from 27 October [[5]](#ref-5). Model Studio's list of models was last updated on 2 November, with Qwen3.8-Max as the newest Max [[6]](#ref-6). Qoder's changelog ends on 9 November with a release about slide templates [[7]](#ref-7).\n\n2026\n\n1. **22 September** (announced). \"its next-generation model, Qwen 4, is currently in training\", at Alibaba's conference in Hangzhou [[2]](#ref-2)\n2. **27 October** (no qwen4). The newest entry in Qwen Cloud's model changelog [[5]](#ref-5)\n3. **2 November** (no qwen4). Model Studio's list of models is last updated, Qwen3.8-Max the newest Max [[6]](#ref-6)\n4. **9 November** (price). Token Plan's monthly credits go up and Pro's price comes down [[3]](#ref-3)\n5. **11 November** (preview). Qwen4-Max in preview on Token Plan, by a post on X [[1]](#ref-1)\n6. **11 November** (no qwen4). No model id, price or limits on any page of Alibaba's I could read [[3]](#ref-3) [[5]](#ref-5) [[6]](#ref-6)\n\nI read all of these within an hour or so of the post, so some of them will have caught up by the time you read this. The press got further than the docs. AlphaSignal found the ids `qwen4-max`, `qwen4-plus` and `qwen4-coder-plus` in a pull request to Qwen Code and a default of a million tokens in the client, and said itself that these are client settings and that Alibaba has published no benchmarks for the model [[8]](#ref-8).\n\n## What the preview is\n\nI don't mind a preview inside a subscription. Alibaba did the same with Qwen3.8-Max, whose preview id is now retired and routes to `qwen3.8-max` [[3]](#ref-3), and I think it's a fair way to find out how a model behaves before you put a price on it.\n\nThe terms are the part to read before you plan anything around it. Token Plan is for interactive use inside programming and agent tools, and it prohibits automation scripts, custom application back-ends and non-interactive batch calls, with suspension or a banned key for use outside that [[3]](#ref-3). It runs in Singapore only, prompts and outputs cross borders, and an individual plan can't be cancelled once bought [[3]](#ref-3). So you can sit in Claude Code this week and point it at Qwen4-Max, and you can't put Qwen4-Max behind a product or leave an agent calling it from a script.\n\nI don't count a model as released until a script can call it and I can read what it costs. By that count Qwen4-Max isn't released yet.\n\n## Why it isn't on Anchor Terminal yet\n\nWe list 29 of Qwen's open-weight models and Qwen Code, and Qwen 4 has been on [Coming models](/upcoming/) since October. This morning it moved from announced to preview there, with the three pages I read. It gets a listing of its own the day one of Alibaba's pages has its id and its price, and the grade will come from those pages and not from the post.\n\n## What I'd like to see\n\nA model card. The id in the docs' list, and a price per million tokens on the pay-as-you-go API, so an agent that isn't a person at a keyboard can use it. Benchmarks someone else can run. And a word on whether any of Qwen 4 will be open weights, which September's release didn't say [[2]](#ref-2).\n\nAlibaba's post asks people to put Qwen4-Max to the test on a real problem. If you have a Token Plan and you do, I'd like to read what you find, and I'll be reading the docs again tomorrow.\n\n---\n\nMore in [Opinion](https://www.anchorterminal.com/blog/opinion/index.md)\n\n- Older: [Google made four TPUs this year, and the rest of us get an interest form](https://www.anchorterminal.com/blog/google-tpu-8-four-chips-and-an-interest-form.md)\n- All posts: https://www.anchorterminal.com/blog/index.md\n\nAnchor Terminal grades the tools AI agents run on, with a source for every score. [The top tools](https://www.anchorterminal.com/top/index.md) · [Categories](https://www.anchorterminal.com/categories/index.md) · [Compare two tools](https://www.anchorterminal.com/compare/index.md) · [Prices](https://www.anchorterminal.com/prices/index.md) · [Sunsets](https://www.anchorterminal.com/sunsets/index.md) · [API, MCP server and CLI](https://www.anchorterminal.com/agents/index.md)\n\n## References\n\nLinks checked on 11 November 2026.\n\n[1] Alibaba's Qwen team (@alibaba_qwen), post on X, 11 November 2026, 02:01 UTC. Source for \"Qwen4 is coming\", the claims of deeper knowledge, richer world understanding and stronger coding and agentic capabilities, Qwen4-Max being in preview on Token Plan, early access on QwenWork, Qoder, the Qwen app and QwenStudio, and the request to test it on a real problem. https://x.com/alibaba_qwen/status/2120335993643684315\n\n[2] Alibaba Cloud, Alibaba Unveils Roadmap on Full-Stack AI Strategy from Chips, Cloud Infrastructure, Models to Agents, 22 September 2026, Hangzhou. Source for \"its next-generation model, Qwen 4, is currently in training\", and for the release naming no Qwen 4 variants and no open weights. https://www.alibabacloud.com/press-room/alibaba-unveils-roadmap-on-full-stack-ai-strategy\n\n[3] Qwen Cloud, Token Plan individual, documentation. Source for the four plans, their prices, credits and concurrent agents, the $3 price for new customers until 30 November 2026, the change to credits and Pro's price on 9 November, how credits are worked out, the list of models ending at qwen3.8-max, qwen3.8-max-preview being retired, the terms on interactive use, the Singapore region and no cancellation. https://docs.qwencloud.com/token-plan/personal/token-plan-personal-overview\n\n[4] Qwen Cloud, Token Plan, pricing page. Source for the tools it works with and the OpenAI and Anthropic protocols. https://www.qwencloud.com/pricing/token-plan\n\n[5] Qwen Cloud, model changelog. Source for the newest entry being dated 27 October 2026 and for no Qwen4 model on the page. https://docs.qwencloud.com/changelog/models\n\n[6] Alibaba Cloud Model Studio, Supported models and capabilities overview. Source for the page being last updated on 2 November 2026, qwen3.8-max as the newest Max, and no Qwen4 model. https://www.alibabacloud.com/help/en/model-studio/models\n\n[7] Qoder, changelog. Source for the newest entry, Qoder 0.5.7 of 9 November 2026, and for no mention of Qwen. https://changelog.qoder.com/app\n\n[8] AlphaSignal, Alibaba Launches Qwen4-Max With Agent-First Coding and 1M Token Support, 11 November 2026. Source for what it reported, the ids qwen4-max, qwen4-plus and qwen4-coder-plus in a Qwen Code pull request, a default of a million tokens in the client, and that these are client settings with no benchmarks published by Alibaba. https://alphasignal.ai/news/alibaba-launches-qwen4-max-with-agent-first-coding-and-1m-token-support\n",
  "meta": {
    "attribution": "Anchor Terminal (https://www.anchorterminal.com)",
    "attributionRequired": true,
    "credit": "CC BY 4.0, credit Anchor Terminal and link the page or API address you took it from (https://www.anchorterminal.com/terms/#data)",
    "docs": "https://www.anchorterminal.com/docs/",
    "generatedAt": "2026-11-11",
    "license": "CC-BY-4.0",
    "method": "https://www.anchorterminal.com/benchmark/",
    "methodology": "0.7",
    "openapi": "https://www.anchorterminal.com/openapi.json",
    "preview": false,
    "run": "2026-10-01",
    "runLabel": "October 2026 research run",
    "terms": "https://www.anchorterminal.com/terms/#data"
  },
  "page": {
    "breadcrumbs": [
      {
        "name": "Home",
        "url": "https://www.anchorterminal.com/"
      },
      {
        "name": "Blog",
        "url": "https://www.anchorterminal.com/blog/"
      },
      {
        "name": "Opinion",
        "url": "https://www.anchorterminal.com/blog/opinion/"
      },
      {
        "name": "Qwen4-Max in preview",
        "url": ""
      }
    ],
    "description": "Vlad Cealicu on Qwen4-Max, which Alibaba's Qwen team put into preview on its Token Plan subscription on 11 November 2026. It can be tried in a coding tool from $6 a month, but no page of Alibaba's names its id, its price or its limits yet, and the plan's terms rule out scripts and back-ends.",
    "facts": [
      "Vlad Cealicu",
      "11 November 2026"
    ],
    "h1": "Qwen4-Max is in preview, and Alibaba's own docs still stop at Qwen3.8",
    "image": "https://www.anchorterminal.com/assets/og/blog-qwen4-max-preview-docs-still-say-3-8.png",
    "path": "/blog/qwen4-max-preview-docs-still-say-3-8",
    "published": "2026-11-11",
    "section": "blog",
    "title": "Qwen4-Max preview on Token Plan, and what the docs say | Anchor Terminal",
    "toc": null,
    "updated": "2026-11-11",
    "url": "https://www.anchorterminal.com/blog/qwen4-max-preview-docs-still-say-3-8"
  },
  "tokens": {
    "markdown": 2350,
    "slim": 930
  },
  "version": 1
}