Alibaba’s Qwen3.7 Preview Tops China in Text and Vision Rankings

Alibaba’s Qwen team has quietly rolled out the first preview of the Qwen3.7 series on chat.qwen.ai, introducing two variants: Qwen3.7-Max-Preview and Qwen3.7-Plus-Preview. Both models are forcibly locked into deep-thinking mode, temporarily disabling web search and code interpreter functionalities—a move that narrows the model’s utility but focuses it purely on reasoning. The preview arrives just 28 days after Qwen3.6-Max-Preview, an astonishingly fast product iteration cadence that underscores Alibaba’s aggressive pace in the large-model wars. The timing is strategic: the launch sets the stage for the 2026 Alibaba Cloud Summit opening on May 20 in Hangzhou, where official technical foundations and commercial deployment plans for the base model will be unveiled.

In the latest Arena leaderboard, Qwen3.7 has achieved dual top rankings in China for both text and vision capabilities. Qwen3.7-Max-Preview sits at 13th place globally in text, elevating Alibaba Lab’s text research ranking to 6th worldwide. More impressively, the model enters the global top 10 across three core reasoning benchmarks: 7th in mathematics, 9th in expert prompting and software/IT, and 10th in code generation. On the vision side, Qwen3.7-Plus-Preview secures 16th globally, pushing Alibaba’s overall vision research capability to 5th worldwide. This dual strength—text and vision in the same family—positions Qwen3.7 as a competitive multimodal contender against models from OpenAI, Google, and Anthropic. For professionals, it means another high-performance open-source-aligned alternative for reasoning-heavy tasks and multimodal applications, with enterprise deployment details expected at the summit.