Resolves YES if, by November 25, 2026, 23:59 UTC, Alibaba's Qwen team publishes downloadable model weights for any model officially named Qwen 4 (any size or tier, e.g. "Qwen4-27B", "Qwen4-Flash") on its official Hugging Face organisation, https://huggingface.co/Qwen. The weights must be public: anyone with a free Hugging Face account can download them, even if the repo asks you to accept a licence first.
Resolves NO otherwise.
What doesn't count:
API-only or chat-app-only releases (Qwen 4 Max on Alibaba Cloud / QwenCloud with no downloadable weights).
Models named Qwen 3.x, even if Alibaba describes them as a "preview of the Qwen 4 architecture" (for example the Qwen3.8-Flash preview released on Aug 24, 2026).
Weights uploaded only by third parties (re-uploads, quantisations or leaks), or weights that appear only on ModelScope.
Resolution source: the huggingface.co/Qwen model list and each repo's commit history (the date of the first public commit containing weight files). I'll post a link to the repo when resolving.
Background: At the Apsara Conference on Sep 22, 2026, Alibaba previewed four Qwen 4 tiers: Max, Plus and Flash as hosted API models, plus a 27B open-weights model. They said Qwen 4 was "in training" and would arrive "very soon". As of Oct 24, 2026, no release date has been announced and no Qwen 4 weights are on Hugging Face.
I won't trade in this market during its first 7 days.
1,000I've seeded this at 42%. Alibaba said on Sep 22 that Qwen 4 was "very soon" and named a 27B open-weights tier. Qwen has usually put out open weights the same day as its API models. But a month has gone by with no date, and the lab only does one big launch at a time, so a 4-week window is close to a coin flip.
Sources: https://cellcog.ai/blog/qwen-4-release-date/ (Apsara, "in training", no date) and https://huggingface.co/Qwen (no Qwen 4 repo yet).
What would move me: a Qwen4 pull request merged into transformers or vLLM (I'd go up a lot), or Alibaba launching Qwen 4 Max as API-only with no word on weights (I'd go down).