Where it stands

Status: Spotted. First seen 10 Nov 2026 in photo of JavaScript source excerpt, Qwen Chat build 0.3.40. 1 sighting.

Leak@LuminaBench

LuminaBench Posts Photo of Qwen4 Max Preview ID in Qwen Chat Code

LuminaBench posted a photo of a JavaScript excerpt from Qwen Chat build 0.3.40, where the agent mode model ID changes from qwen3.8-max in build 0.3.39 to qwen4.0-max-preview. The photo is the post's only evidence for the change. ArtificialWatch reports that as of Nov 8 it found no Qwen 4 model ID on Hugging Face, OpenRouter, models.dev or Qwen Chat, and that Qwen Chat's list then included Qwen3.7-Plus, Qwen3.8-Max and Qwen3.8-Omni-Flash. ArtificialWatch also reports that Alibaba said on Sep 22 at its Apsara Conference that Qwen 4 is in training, with no release date. Qwen Code's Oct 28 rule gives qwen4-prefixed model IDs a 1M-token default context, which ArtificialWatch describes as a tool setting rather than a model spec.

Image shared by @LuminaBench on X
Image shared by @LuminaBench on X

Sources: artificialwatch.com, our earlier report.

All tracked models →

Source