Alibaba's next flagship model is here — for two weeks only. Qwen4-Max-Preview opened on November 11 as the first public model of the Qwen4 generation, and Alibaba has been unusually explicit about the terms: the preview runs until November 25, the model will not change during that window, and it will be retired when the window closes. It is a launch with an expiry date printed on the box.

What actually shipped#

Qwen Cloud's documentation describes Qwen4-Max-Preview as the next-generation flagship multimodal model in the Qwen family, built for coding, office productivity, front-end design and complex agentic tasks. It accepts text, images and video, and returns text. The context window is one million tokens, and the maximum output is 393,216 tokens — shared between the model's thinking and its final answer, so long reasoning chains eat into the response budget. Thinking is on by default. The model is compatible with OpenAI's Chat Completions and Responses APIs and Anthropic's Messages API, and it supports tool calling, which means existing agent stacks can point at it with minimal rewiring. The official Qwen account announced it on X on November 11 at 02:01 UTC.

The two-week window#

Access is the real story of this launch. Qwen4-Max-Preview is available only through Alibaba's Token Plan, a monthly subscription for coding and agent tools; pay-as-you-go API access is not supported, and calls go to a dedicated base URL. The model also appears in Qwen Studio, the Qwen app, QwenWork and Qoder. Alibaba is running a promotion during the preview that cuts Token Plan Credits consumption to as little as 20% of the standard rate, according to Asia Daily — a discount that reads as an invitation to test hard while the window is open.

Alibaba Group headquarters in Hangzhou, with its lattice-patterned facade
Alibaba Group headquarters, Hangzhou. Photo: Thomas LOMBARD, CC BY-SA 3.0, via Wikimedia Commons.

What Alibaba didn't publish#

What the launch omits is as telling as what it includes. At launch Alibaba had published no benchmark table, no model card, no parameter count, no license terms and no per-token price — so every quality claim about Qwen4-Max is, for now, the Qwen team's own word. There are no weights either: nothing named Qwen4 exists in the Qwen organisation on Hugging Face as of November 11, and the preview runs only on Alibaba's servers. That is a sharp break from the pattern of the last generation — Qwen3.8-Max, the previous flagship, launched on August 3, 2026, and Alibaba published its weights as Qwen3.8-2.4T-A95B shortly after. The rest of the Qwen4 family — Qwen4-Plus, Qwen4-Flash and a 27B model intended for local deployment — has no release date.

A developer's desk at night with a laptop showing code and a large glowing hourglass beside it
AI-generated editorial illustration: the two-week preview window, counting down.

Why a timed preview#

The format makes sense as a controlled trial. A frozen, time-boxed preview lets Alibaba gather usage and feedback on a flagship without committing to support it, price it, or open-weight it — and it funnels developers into the Token Plan subscription. For developers, the implication is equally clear: this is an evaluation window, not a foundation to build on. Anything shipped against Qwen4-Max-Preview needs a migration plan for November 26, when the model ID stops working.

What to watch#

Three questions decide whether this preview is a footnote or the start of the Qwen4 era. First, whether independent benchmarks appear before the window closes — the model's quality is currently unverified by anyone outside Alibaba. Second, whether a full Qwen4 launch follows, and whether it restores the open-weights pattern that made Qwen the default choice for local deployment. Third, whether the 27B local model materialises, which would tell us how much of the Qwen4 generation Alibaba intends to share at all. Until then, Qwen4-Max is a two-week audition: impressive on paper, unproven in public, and already counting down.

Sources: Alibaba Cloud Model Studio — Qwen4-Max-Preview (Nov. 11, 2026); Modelfit — Qwen4-Max Preview: 1M Context, Cloud Only, No Weights (Nov. 11, 2026); Fello AI — Qwen 4 Release Date: Qwen4-Max Preview Is Out (Nov. 11, 2026); Asia Daily — Alibaba Opens Qwen4-Max Preview With Million-Token Context (Nov. 11, 2026). No independent benchmarks exist yet; capability claims are Alibaba’s own.