Alibaba has introduced its latest AI model, Qwen 3.7 Max-Preview, on May 14, 2026, ahead of the Alibaba Cloud Summit, placing it as the #13 model globally in text rankings and marking Alibaba as the sixth-ranked AI lab worldwide. This strong performance follows a trend where Chinese tech companies like Alibaba are increasingly seen as viable alternatives to leading Western AI systems, particularly excelling in areas such as math, coding, and multilingual reasoning. Notably, while the Qwen 3.7 Plus variant will be open-sourced to developers, the flagship Qwen 3.7 Max remains proprietary, reflecting Alibaba’s strategy to monetize its top models while providing access to slightly lower-tier options.
Qwen: Qwen is Alibaba’s family of large language models, designed for tasks like coding, multilingual reasoning, and creative writing, and has quickly climbed independent leaderboards that compare cutting-edge AI systems. In this article, the new Qwen 3.7 Max-Preview and Qwen 3.7-Plus-Preview are evaluated against rivals like Xiaomi’s MiMo, highlighting strengths in math and structured reasoning, solid but less expressive creative output, and a split strategy where the Plus model is open source but the higher-performing Max model remains closed.
Alibaba: Alibaba is a major Chinese technology conglomerate whose cloud division has become a key player in developing large language models under the Qwen brand. In this news, Alibaba is using the Qwen 3.7 Max and Plus previews to showcase rapid progress toward frontier-level AI performance, while signaling a business strategy that keeps its most capable models proprietary and monetized through Alibaba Cloud.
Arena AI: Arena AI is an evaluation platform that runs blind, crowd-sourced comparisons of large language models across tasks like text generation, coding, and reasoning, producing public leaderboards that many labs use as an external benchmark. In this story, Qwen 3.7 Max-Preview and Plus-Preview debut on Arena AI’s Text Arena, where Qwen 3.7 Max’s strong ranking is used by Alibaba both to validate the model before launch and to demonstrate that its capabilities are approaching those of leading Western labs.
China_AI_Competition: Chinese tech companies such as Alibaba and Xiaomi are increasingly positioning their frontier models as viable alternatives to leading Western systems, with recent independent tests emphasizing strengths in math, coding, and multilingual reasoning.
Open_Source_Strategy: Alibaba’s decision to open source the Qwen 3.7 Plus variant while keeping Qwen 3.7 Max proprietary reflects a broader trend among major labs of releasing strong but not flagship models to developers while reserving their best systems for paid cloud access.
Evaluation_Methodology: Arena AI’s blind, head-to-head comparisons have become an influential way for the AI community to assess model quality based on real user preferences rather than static benchmark scores.