Alibaba opens Qwen4 flagship preview for two weeks, with no date for the full release

Alibaba opened a limited preview of Qwen4-Max, the flagship of its next-generation Qwen4 model family, on November 11. The preview lasts two weeks and the model will be retired after November 25; the full Qwen4 release and the other tiers still have no announced date, and Alibaba has published no benchmark scores.

What is in the preview

The official Qwen account posted “Qwen4 is coming” at 10:00 a.m. Beijing time on November 11, announcing that Qwen4-Max-Preview (model ID qwen4-max-preview) was live on the Token Plan subscription, and available the same day in Qwen Chat as well as QwenWork, Qoder, the Qwen app and QwenStudio. Alibaba's documentation gives the preview model a 1,000,000-token context window and an output ceiling of about 393,216 tokens, shared between reasoning and the final answer, with text, image and video input and thinking on by default. The preview runs from 10:00 on November 11 to 24:00 on November 25, Beijing time, after which the model will be retired, with no changes during the window.

Narrow access

The preview is available only to Token Plan subscribers. Pay-as-you-go billing is not supported, the API runs through a dedicated Singapore endpoint, and an ordinary Model Studio key cannot call the model. Alibaba has published no parameter count, released no weights, set no per-token price and reported no benchmark score; the Artificial Analysis leaderboard still carries no Qwen4 entry.

Full release undated

The Qwen4 family was first shown on September 22 at the Apsara conference in Hangzhou, when Alibaba named four tiers — Max, Plus, Flash and a 27B open-weights model — but only the Max preview has shipped so far, and the rest of the family and the full release have no date. Qwen3.8-Flash-Next, open-sourced in late August, was described by Alibaba as an experimental preview of the Qwen4 architecture. For comparison, the previous flagship, Qwen3.8-Max, has been generally available since August 3 with the same 1,000,000-token context, priced at $2.00 per million input tokens and $6.00 per million output tokens.

Sources: Alibaba Qwen official account and documentation; Yotta Labs; Cellcog.