the latest-generation large models from alibaba’s qwen, qwen3.7-max-preview and qwen3.7-plus-preview, have officially launched on the qwen chat and arena ai platforms (formerly known as lmarena), open for global users to experience. according to official sources, this model series is expected to make its world debut during the alibaba cloud summit on may 20.
as a preliminary version of the qwen3.7 flagship series, qwen3.7-max-preview focuses on exceptional reasoning capabilities, demonstrating strong performance across multiple authoritative benchmarks. currently, both preview models are in “thinking mode” and do not yet enable advanced features such as internet search or code interpreters. on the arena ai text comprehension leaderboard, qwen3.7-max-preview ranks 13th, while in alibaba’s qwen self-built laboratory evaluation system, it holds the 6th position. its performance across specific dimensions is particularly noteworthy: mathematical reasoning ranks 7th, expert-level applications and the software/it domain both rank 9th, and programming ability also places it within the top 10.
in visual understanding, qwen3.7-plus-preview ranks 16th on the arena ai visual comprehension leaderboard, helping qwen climb to 5th place among all laboratories participating in visual evaluations. notably, in the specialized “expert arena” test—designed to assess complex prompt engineering skills—qwen3.7-max-preview also secured 9th place. as the alibaba cloud summit approaches, industry observers are closely watching the full evolution path of the qwen3.7 series in areas such as multimodal collaboration, tool invocation, and production‑grade deployment.
