Qwen Office's recently expanded model selection has been quickly rolled back.

On the evening of August 17, it was discovered that the "frontier models" — Z.AI GLM-5.3 and DeepSeek V4 Pro — which had been made available on Qwen Office, have disappeared from the product interface. The available tiers have reverted to "Premium," "Basic," "Economy," and Qwen3.8-Max.

This change comes just three days after the two external models were introduced. On the evening of August 14, Qwen Office had announced the initial integration of GLM-5.3 and DeepSeek V4 Pro. Combined with the previously available Qwen3.8-Max, its frontier model lineup had briefly covered Qwen, Z.AI, and DeepSeek.

At that time, Qwen Office emphasized that it would maintain this integration pace consistently. However, as of the evening of August 17, the WeChat official account article announcing the launch has been restricted to the author's privacy settings, rendering it publicly inaccessible.

Alibaba has yet to provide a public explanation for the removal of the two models. Consequently, it remains uncertain whether this adjustment signifies a short-term product change or a shift in Qwen Office's overall strategy for integrating external models.

Nevertheless, this brief launch and swift withdrawal reveals a subtle internal dilemma. For Qwen Office, a larger pool of external models means richer capabilities it can leverage for its product. Conversely, for Qwen, real-world tasks serve as a crucial source of usage and feedback.

Balancing the goal of building a superior product against the need for its proprietary model to gain access to more genuine tasks, Alibaba must weigh both considerations simultaneously.

With competitors like Tencent's Workbuddy and ByteDance's Trae Work moving toward multi-model aggregation, Alibaba still faces a fundamental question: Is Qwen Office intended to be a more open office agent, or a stronger Qwen-powered office scenario?