API first, weights later: the sequence that now defines an open-weight release
Qwen3.8-Max went live through Alibaba Cloud's Model Studio with published specs and API pricing, and open weights described as following afterwards. Every lab running both an open and a commercial line now ships in that order, and the gap between the two dates is where the meaning of open weights is quietly being decided.
Nobody is being dishonest about this. The release is real, the specs are published, and the weights are said to be coming. But a reader who tracks the label rather than the calendar forms a steadily less accurate picture of what is actually runnable today versus what is merely announced.
The lag matters most for the capabilities that are hardest to reproduce independently. A million-token context window is expensive to serve and difficult to run locally, so it is exactly the sort of feature where an API-first period leaves self-hosting users furthest behind — and it is the axis they can least compensate for with ingenuity.
Which is why the gap is worth watching as a metric in its own right. Not whether a lab calls itself open, but how many weeks separate the endpoint from the download, and whether that number is growing. Meta's Behemoth has been in that gap since April 2025 and never emerged.
TestingCatalog — Alibaba released Qwen3.8-Max with open weights coming soon → · Briefs — Qwen3.8-Max launch lifts Alibaba stock → · Fello AI — Qwen 3.8: Alibaba's 2.4T model →