New on the radar: Qwen 4 Max appeared in chat.qwen.ai (first-party app)

Our hourly sweep found an id that was not there an hour ago.

Catalog id: qwen4.0-max-preview
First observed by our sweep: 2026-11-11 03:43 UTC
Source: chat.qwen.ai (first-party app)

This entry is machine-written from what the sweep actually read, and nothing else. A catalog listing is a primary source for exactly one narrow fact — that this id now exists, at this price, with this context window — and it is not evidence of anything beyond that. It does not tell you the model is good, that the listing is intentional, that the weights are open, or that the vendor has announced it. Those questions need a human and an editor's judgement, and this item deliberately does not answer them.

It maps to Qwen 4 Max, already on our watchlist — registry note: no date · Alibaba says Qwen 4 is in training (Apsara, Sep 22) · Oct 22: @LuminaBench and @Mr_Salio say Qwen 4 is being tested on LMArena under the codename 'Kalmia', possibly in more than one version (unconfirmed; Alibaba has said nothing, and no Qwen 4 id is in OpenRouter or models.dev) · Oct 26 09:34 UTC: @Mr_Salio posts a screenshot of a code list naming 'qwen4-exp' next to 'qwen3.8-flash-next', without saying where it comes from, and expects Qwen 4 this week or next (unconfirmed; our hunt fires on qwen4-exp) · Oct 27 10:16 UTC: @Mr_Salio reports another Qwen 4 checkpoint on LMArena, codenamed 'Kerria' (unconfirmed) · Oct 29: at Alibaba Cloud's first international Apsara Conference in Amsterdam (the event is confirmed by Alibaba Cloud's own posts), attendee @jtdavies quotes Alibaba Cloud CTO Feifei Li on stage: 'Qwen4 next week' (08:57 UTC), then posts 'Qwen4, coming soon' from a talk by Qwen research scientist Xudong Guo (09:20 UTC); @scaling01, @Mr_Salio and @LuminaBench relayed it · one attendee's account, and Alibaba's own posts from the event so far cover the new Netherlands cloud region, not Qwen 4; no Qwen 4 id in OpenRouter or models.dev (read 11:00 UTC) · Oct 28: Alibaba's open-source Qwen Code repo merges a rule sizing every qwen4 id at 1M tokens as a 'forward default', with tests naming qwen4, qwen4-max, qwen4-plus and qwen4-coder-plus (PR #15762, read directly; the 1M is the tool's default, not a published spec), and a qwen4-omni-120b-think id has sat in its test fixtures since Sep 16; first flagged by @Mr_Salio · still no Qwen 4 id in OpenRouter or models.dev (read 12:00 UTC) · Oct 31 13:43 UTC: @Mr_Salio reads the 'qwen-latest-series-invite' ids in Qwen Code as invite-only Qwen 4 testing ahead of a public release next week (unconfirmed). What the repo shows: invite-only qwen-latest-series-invite-* aliases sized at 1M since Oct 28 (PR #15758, example id 'qwen-latest-series-invite-202610-m4') and grouped with qwen4* for reasoning effort on Oct 31 (PR #15987), read directly; the alias names no generation, and third-party wrappers' docs map earlier invite ids of this form (beta-v16, beta-v24) to Qwen3.7 previews · Nov 4: halfway through the Nov 2-8 window that 'next week' implies, no qwen4 id is on Hugging Face (Qwen's newest repo is Qwen-Image-2.1-Turbo, Oct 9), OpenRouter or models.dev, and Qwen Code has merged no new qwen4 change since Oct 31; Polymarket's 'next Qwen Max (3.9+)' market, which a Qwen 4 Max would settle, last traded at 94% by Dec 31 · Nov 6 12:00 UTC: @Alibaba_Qwen posts a teaser, 'The becoming begins.', with an image of a flock of birds over the sea and the Zhuangzi line 北冥有鱼 其名为_ (the fish Kun that becomes the bird Peng, its name left blank); the post and image name no model and no date, and a widely shared 'Qwen 4' graphic is not from Alibaba's post · @Mr_Salio ('Qwen4-27B is dropping today'), @LuminaBench and TestingCatalog read it as Qwen 4, with Max, Plus, Flash and an open-weight 27B expected and the 27B first (unconfirmed) · at 20:56 UTC no qwen4 id is on OpenRouter, models.dev, Qwen's Hugging Face page or chat.qwen.ai (top model there: Qwen3.8-Max), and Qwen Code has merged no qwen4 change since Oct 31 · the RSI result Alibaba published (Sep 22) is on Qwen3.8-Max (33 automated cycles, Artificial Analysis score 40 to 45); the release does not say how Qwen 4 uses RSI · the hunt is split Nov 6: qwen4 Flash and 27B ids go to their own rows, and all three rows are headline, so a Qwen 4 launch reaches more than the all-models list · Nov 8: qwen4 Plus ids get their own headline row too, and qwen4 Coder, VL, Omni and small-size ids no longer fire this Max hunt · Nov 10: the Nov 2-8 window that the CTO's 'next week' implied has now PASSED with nothing shipped — Alibaba has still announced no Qwen 4 and no date, and our sweep's cumulative ledger of 639 catalog ids has never once carried a qwen4 id · Nov 10: Qwen Chat's own front-end code (build 0.3.40 on Alibaba's CDN, served by chat.qwen.ai, read 19:52 UTC) makes 'qwen4.0-max-preview' the default model for agent mode in place of qwen3.8-max (build 0.3.39) and adds an 'agent_qwen4' announcement pop-up ('Try in Chat'); flagged by @LuminaBench (19:22 UTC) · chat.qwen.ai's public models list still tops out at qwen3.8-max and OpenRouter has no Qwen 4 id; this hunt fires on qwen4.0-max-preview · Nov 11: Qwen4-Max-Preview (qwen4.0-max-preview) is live in Qwen Chat: first seen on chat.qwen.ai's public model list at 02:10 UTC (radar variant alert 02:11 UTC), read directly 02:47 UTC; 1,000,000-token context, vision, document, video, audio, thinking and search flagged; no API (Model Studio tops out at qwen3.8-max), not on OpenRouter, no price and no Alibaba post yet · the preview fires the variant stage only, so this row stays armed for the final Qwen 4 Max.