Alibaba announced Qwen3.7-Max on 21 May 2026 at the Alibaba Cloud Summit. It is a proprietary, closed-weight, API-only reasoning and agent model with a 1 million token context window. The release marks a shift away from Qwen’s previously open-weight flagship models.

What happened

Alibaba announced Qwen3.7-Max on 21 May 2026 at the Alibaba Cloud Summit.

Qwen3.7-Max is a proprietary, closed-weight, API-only reasoning and agent model. It offers a 1 million token context window and is served through Alibaba Cloud Model Studio. The launch marks a shift from the previously open-weight Qwen flagship models.

Why it matters for builders

The move to a closed-weight, API-only flagship changes how teams adopt Qwen at the top tier. Where earlier flagship models could be downloaded and self-hosted, Qwen3.7-Max is reached only through Alibaba Cloud Model Studio, which ties frontier access to the platform.

As a /glossary/reasoning-models/ built for agent workloads, Qwen3.7-Max targets multi-step planning and tool use. Its 1 million token context window makes it a /glossary/long-context-model/, able to hold long documents and extended agent state in a single request. Teams weighing open versus closed options should note that the open-weight path no longer covers Qwen’s top model. See /tools/alibaba-qwen/ for access details.

Sources

  1. Alibaba / Qwen. “Qwen3.7: The Agent Frontier” (21 May 2026). https://qwen.ai/blog?id=qwen3.7