Qwen 4 is starting to look close. We spotted preparations in Qwen Code for both Qwen4 model names and an invitation-only preview family, with reasoning controls being added for them. 1 2
The capture names qwen4-plus and Qwen4-Max, alongside qwen-latest-series-invite-202610-m6. They get Low, Medium and xHigh reasoning options, with xHigh selected by default and an option to turn thinking off. 2
What caught our eye is the work around the names. Qwen Code is being taught to recognise the new families and show the appropriate controls, and the changes include tests for those choices. That looks like the client being made ready for a rollout. 2 3
The author says the checks used built packages and request tests, without live model calls. We have not used Qwen 4 through this change. 2
Our expectation is a release next week, in the week beginning 2 November. That is our read of the preparations; the PR itself does not announce a launch date. 1
Sources & references
- Lumina: Qwen 4 release preparations ↗
31 October 2026. Owner-supplied Lumina finding and original capture. The next-week release window is Lumina’s expectation.
- Qwen Code: reasoning controls for new Qwen model families ↗
Checked 31 October 2026. The PR description names the invitation-only preview and qwen4 families, lists the available effort settings and says no live model calls were made.
- Qwen Code: model-configuration test changes ↗
Checked 31 October 2026. The test diff includes qwen4-plus, Qwen4-Max-2026-10-01 and the invitation-only preview identifier. These are configuration fixtures, not proof of public model availability.
