If a preview model powers a demo this week, its retirement date belongs in the deployment plan.
Alibaba Cloud opened Qwen4-Max-Preview on November 11. The model accepts text, images, and video, and Alibaba lists a one-million-token context window. API access requires a Token Plan subscription and a dedicated key for its Beijing endpoint. The preview has no pay-as-you-go access. Alibaba says the model ID will retire at midnight at the end of November 25 in UTC+8, which is 16:00 UTC that day.
Check what the output limit counts
On the OpenAI-compatible Chat Completions route, Alibaba says max_tokens and max_completion_tokens both cap reasoning and the final answer together. If your application budgets 2,000 tokens for a report, the model can spend part of that allowance reasoning before it writes the report. This behavior is documented in the API compatibility notice.
Take one representative long task that returns structured data. Run it with your existing cap and with a larger cap, then compare whether both responses complete and satisfy your schema. Repeat the same test with the fallback model. Keep the provider endpoint and model ID configurable so the change can be rehearsed before the preview ends. Alibaba’s documentation establishes the preview deadline. It gives no date for a successor.
Abdessalam Alaoui
I'm Abdessalam Alaoui, founder of Musthave.AI. I write practical guides to AI models, tools, generative media.
