Alibaba retires over 100 legacy Qwen model IDs today

Alibaba Cloud’s Model Studio retires more than 100 legacy model IDs today, 10 October. The list spans the Qwen lineup, including qwen-turbo, qwen-vl-max and older qwen3-max snapshots, plus some third-party models it hosts, such as older DeepSeek, GLM-4 and Moonshot Kimi variants.

Alibaba’s notice is blunt about what happens next: “any applications or services continuing to call them will fail to receive responses.” There’s no redirect and no graceful error, just silence. The date itself was already pushed back once, from an earlier deadline, to “align the model retirement schedule with user migration timelines.”

Why it matters: if you’ve got one of these model IDs hardcoded, for example qwen3-max-2025-09-23 or an old qwen3-vl-flash snapshot, requests stop working today, not with a deprecation warning first. Alibaba’s notice lists recommended replacements for each retired model, such as qwen3.7-max for the old max snapshots and qwen3.6-flash for the old VL-flash tier.

The caveat: Alibaba’s own notice flags the date as “subject to the actual change time”, so some endpoints may linger a little past today. That’s not something to plan around.