Alibaba announced Qwen 3.8 on July 19, 2026, but the full model has not shipped on a public API. The model is reportedly 2.4 trillion parameters and slated to ship with open weights. A hosted Qwen3.8-Max-Preview is already running on Alibaba's Token Plan, Qoder, and QoderWork, with the open-weight release expected to follow the preview. No firm public date has been given; the article describes the release as "imminent but unconfirmed." To prepare, the article recommends keeping integration model-agnostic by using a router like OrcaRouter through one OpenAI-compatible endpoint, which exposes the Qwen 3.8 model ID today and returns a not-yet-available status until launch, then begins succeeding automatically. Builders can test against Qwen 3.7 Max or the preview as stand-ins, keeping the model name configurable so switching is a one-line change.