By Hamza Faisal
Alibaba has broken its silence on Qwen 4 — with a single sentence. On Friday, November 6, the official Qwen account on X posted one line, "The becoming begins.", alongside a poster and no further explanation. Within 14 hours the post had passed 1.6 million views, and the open-source AI community had read it as the stealth announcement of Alibaba's next flagship model family.
There was no keynote, no livestream and no benchmark chart. Alibaba has still not published a release date, a model card, a price or a single specification for Qwen 4. As of Saturday morning, no Qwen 4 repository exists on Hugging Face. But after months of stage teases, arena sightings and code leaks, the cryptic post is the closest thing to an official confirmation that the launch is now imminent.
A poster full of clues, and a blank where the name should be
The poster's artwork carries the message. It shows a flock of birds shaped like a fish flying over the sea at dusk, above the opening of a line from the Zhuangzi, the ancient Chinese text: "In the northern darkness there is a fish, and its name is ___" — with the name left blank. In the classic story, the fish is Kun, which transforms into the giant bird Peng. Readers took the blank as a name still to come.
The tease caps a runway that has been building for months. At Alibaba's Apsara Conference in Hangzhou on September 22, Qwen project lead Liu Dayiheng confirmed that Qwen 4 was in training on a new-generation architecture and would arrive "very soon." Then, on October 29, at Alibaba Cloud's first Apsara Conference in Europe in Amsterdam, a stage slide read "Qwen4 Series Coming Soon" and named four tiers: Qwen4-Max, Qwen4-Flash, Qwen4-Plus and Qwen4-27B. Alibaba Cloud chief technology officer Feifei Li reportedly told attendees the launch would come "next week" — a window that runs out this weekend.
Technical preparation has been visible in public for weeks. In late August, Alibaba open-sourced Qwen3.8-Flash-Next, a 125-billion-parameter mixture-of-experts model, and described it on its model card as an early preview of the architecture that will underpin Qwen 4. Repository updates have since assigned Qwen 4 identifiers a context window of one million tokens.
Why this launch matters beyond the tease
Qwen is already one of the most widely used open model families in the world, with more than 470 open-source variants, and its next generation arrives at a moment of unusual attention for Alibaba's AI business. Just last week, Apple confirmed that Qwen will power Apple Intelligence in China, with testing starting on November 4 — news that lifted Alibaba's stock by 5.7 percent. A strong Qwen 4 would give Alibaba a fresh flagship to pair with that high-profile deployment.
It would also land in the middle of a crowded race. OpenAI, Google and Anthropic have all shipped major releases in the past fortnight, and Chinese rival Meituan unveiled its LongCat 2.5 model this week, trained entirely on Chinese-made chips. The open-weight corner of that race is where Qwen has traditionally been strongest, and developers are watching whether Qwen 4 keeps the family fully open.
For now, though, the honest status is narrow: a teaser, a poster and a launch window that is still open. No benchmarks exist for the final Qwen 4 checkpoints, no pricing has been announced, and Alibaba has not confirmed in writing either the four tier names or a date. If the model does land in the coming days, the weights — not the poster — will be the real announcement.
