Qwen3.7-Max

Superseded

Alibaba released Qwen3.8-Max on 3 August 2026 and it now heads the line. The successor also reverses the closed turn described below: Alibaba says it will publish the Max tier’s weights for the first time. The page below stands as the record of where the Max tier sat until then.

What it is

Qwen3.7-Max is the top, Max-tier text model in Alibaba’s Qwen family. It is served through Alibaba Cloud Model Studio as a hosted API, with endpoints across several regions, Beijing, Singapore, Tokyo, Frankfurt and Virginia, and compatibility modes for the OpenAI, Anthropic and DashScope interfaces. In Alibaba’s own line-up it sits at the strongest-capability end of a range that runs from most capable to lowest cost.

The closed turn

The notable thing about the Max tier is what it is not. Qwen built its reputation on open-weight models, families that anyone could download and run, which made it one of the most widely used open model lines in the world. The Max tier breaks with that. Qwen3.7-Max is API-only, its weights closed, available to rent rather than to own. A lab known for openness keeping its most capable model behind an API is a meaningful shift, and it mirrors a wider pattern of Chinese labs releasing capable open weights lower down the range while reserving the frontier tier for paid access.

What is on the record, and what is not

Alibaba documents Qwen3.7-Max as a production text-generation model with the integration details a developer needs, the endpoints, regions and API modes. It has not published the kind of detailed model card, with parameter counts, context window and named benchmark results, that Western labs put out on launch day. That reticence is itself worth noting: the specifications and independent evaluations that would let a buyer compare it precisely with Western flagships are not openly on the record, so any single performance claim should be treated with caution until independently tested.

Why it stands out

Qwen3.7-Max is a marker of where the open-source frontier is heading. Qwen has been a mainstay of the open-weight ecosystem, so the arrival of a closed, API-only Max tier signals that even the most open labs now see their best model as something to monetise directly. What to watch is whether Alibaba keeps releasing strong open weights in the tiers below while the Max line stays closed, and how the Max tier measures up once independent testing catches up with it. For how the Qwen line fits the wider field, see our large language models hub and the AI section.