Qwen3.8-Max Review

Alibaba's flagship, released on 3 August 2026 as the cloud-hosted top of the Qwen3.8 line. Architecturally it is a sparse mixture-of-experts model with about 2.4 trillion total parameters and roughly 95 billion active per forward pass, addressing up to a million tokens of context. It takes text and image input and, in the hosted version, generates text, images and video. Alibaba released open weights shortly after as Qwen3.8-2.4T-A95B on 12 August, minus some cloud-only features, and a much smaller Qwen3.8-27B on 14 August under Apache 2.0 that restores image input for teams running locally.

The licensing on the large model is the detail worth reading closely: providers earning more than fifty million dollars in revenue must negotiate a commercial license with Alibaba, so the open-weights framing has a ceiling on it. Qwen remains the most widely fine-tuned base family outside the Llama lineage. This grade is a provisional placeholder pending hands-on testing.