Alibaba opens Qwen4-Max preview for 15 days — then it disappears
Alibaba Cloud opened Qwen4-Max-Preview on November 11, the first Qwen 4 model anyone can call, with a 1-million-token context window and a 393,216-token maximum output. But the preview runs only until November 25, when Alibaba says the model will be retired — and no parameter count, price, weights, model card or benchmark score has been published.
The News
Alibaba Cloud opened Qwen4-Max-Preview on November 11, the first model from its Qwen 4 family that anyone can actually call. It is the flagship tier: multimodal, taking text, image and video as input, with reasoning and tool calling — the ability to use external tools such as search or code execution while it works. The headline specifications are a 1-million-token context window — the amount of text, images and video the model can hold in mind at once — and a 393,216-token maximum output, shared between the model's internal reasoning and its final answer.
The catch is the clock. Access is Token Plan subscription only, with no pay-as-you-go option, and the preview runs from 10:00 on November 11 to 24:00 on November 25 (UTC+8) — about two weeks — after which Alibaba says the model will be retired. Early access is also open in QwenWork, Alibaba's workplace AI app, Qoder, its coding tool, the Qwen app, QwenStudio and Qwen Chat. The rest of the Qwen 4 family — Plus, Flash and a 27-billion-parameter model — has no date at all.
What Alibaba has not published is just as important: no parameter count, no per-token price, no weights, no model card and no benchmark score. A model card is the document that describes what a model can and cannot do; without one, and without benchmarks, there is no independent way to judge Qwen4-Max against its rivals. Social-media posts calling the model a 'disappointment' have no benchmark to stand on either — there is simply no evidence either way yet.
Qwen4-Max is a flagship with no price, no benchmarks and a fifteen-day expiry date.
Why It Matters
A roughly two-week preview that ends in retirement is an unusual way to launch a flagship. The likeliest reading is that this is a controlled test, not a product launch: Alibaba is letting developers kick the tyres of its most capable model, gathering feedback and usage data, before deciding what the real Qwen 4 release looks like. A model that disappears after two weeks cannot build a developer ecosystem around it, so nobody should be building a business on Qwen4-Max-Preview.
The missing details are the story. Alibaba's Qwen models have a strong reputation — Qwen 2.5 and Qwen 3 were widely used open models — but a flagship with no published benchmarks and no pricing is a black box. Developers choosing a model need to know what it costs per token and how it scores on standard tests; without either, the rational move is to try it while it is free to test and wait for the real numbers. The fifteen-day window also limits what developers can learn: a model that vanishes in two weeks cannot be tested in a real production workload, only in a trial.
For Alibaba, the preview still matters. It signals that Qwen 4 exists and is close enough to show to the public, which keeps Alibaba in the conversation alongside OpenAI, Google and Anthropic at a moment when Chinese models are under intense scrutiny. But a signal is not a product. Until Alibaba publishes benchmarks, pricing and a permanent access route, Qwen4-Max is a teaser — and teasers do not move markets.
Analyst's Take
Our call is Mixed. The first callable Qwen 4 model is real news, and a 1-million-token context window with 393,216 tokens of output is a serious specification. But a roughly two-week preview with no benchmarks, no pricing and a scheduled retirement is a demo, not a launch — and there is no evidence at all to support either the hype or the social-media disappointment around it.
What to watch: whether Alibaba publishes benchmarks or pricing before November 25, and what replaces the preview when it retires. If a permanent Qwen4-Max arrives with competitive scores, the preview will look like a smart soft launch. If it vanishes without a successor, the two weeks will have been a shrug.
Are you bullish or bearish on this story?
By The Numbers
FAQ
What is Qwen4-Max-Preview?
It is the first model from Alibaba's Qwen 4 family that developers can call through Alibaba Cloud — a multimodal flagship that takes text, image and video input, reasons, and can call external tools. It is a limited preview, not a permanent product.
What is a context window?
A context window is how much text, image and video a model can hold in mind at once while it works, measured in tokens — roughly, chunks of words. A 1-million-token window means it can work with very long documents or many images in a single session.
Why does the preview end on November 25?
Alibaba says the model will be retired when the preview window closes at 24:00 on November 25 (UTC+8). It has not said what replaces it, or when the rest of the Qwen 4 family — Plus, Flash and 27B — will arrive.
How much does it cost?
Alibaba has not published a per-token price. Access is through a Token Plan subscription only, with no pay-as-you-go option — so developers cannot simply pay for what they use.
Is the model any good?
Nobody outside Alibaba can say yet: no benchmark score, model card or parameter count has been published. Claims on social media that it is a 'disappointment' have no benchmark behind them either.
👁
Source: Alibaba Cloud