Alibaba did not lead the Qwen3.7-Plus launch with a parameter contest. The model, released on June 2, is positioned around one promise: it can take a task further on its own, from reasoning to tool calls, code writing, test runs and correction loops.
The model accepts images, video and text, but its output remains text. Alibaba describes the package as a multimodal hybrid agent running through Bailian, known overseas as Model Studio, with reinforcement learning from execution feedback and safety controls around automated tool invocation.
The market reaction was immediate. Alibaba shares rose about 6% in U.S. premarket trading to $133.23, while the Hong Kong listing closed up 6.6% at HK$130.90. Qwen3.7-Plus ranked No. 16 on Vision Arena, helping Alibaba reach No. 5 among vision labs, while the related Qwen3.7-Max scored 56.6 on the Artificial Analysis Intelligence Index, described as the highest score for a Chinese model at release.
The important caveat is access. Plus is closed source and API-only for now, with context length, output limits and pricing still undisclosed. Alibaba also signed a six-year exclusive AI and cloud partnership with UEFA that starts in 2027, adding another reason investors read the launch as more than a benchmark update.
Sources: CocoLoop; Alibaba Qwen team; MarkTechPost; GuruFocus; The Standard