Alibaba has released two new AI models, Qwen 3.7 Max-Preview and Qwen 3.7-Plus-Preview, which appeared on Arena AI's leaderboard on May 14. The models are a preview for the upcoming Alibaba Cloud Summit on May 20.
Qwen 3.7 Max-Preview ranks #13 globally in text, making Alibaba the sixth-ranked AI lab worldwide. The Plus variant will be open source, while the Max flagship remains proprietary.
Creative Writing
Qwen 3.7 Max demonstrates strong cultural specificity in creative writing, with a Caribbean time travel story that shows deep understanding of cultural context. However, the writing is more angular and less engaging than some competitors, lacking richness and elaboration as the narrative progresses.
Coding
In coding, Qwen 3.7 Max produced a logically coherent 2D game with thoughtful game design, though it was less visually rich than a competitor's 3D output. The model prefers concise code, which could be a plus for production environments.
Logic and Common Sense
Qwen excelled in logic and common sense, correctly identifying the trick in a question about marriage law and providing a complete, honest response that addressed both the literal and intended interpretations.
Math
This is Qwen 3.7 Max's strongest area. It successfully solved a complex Dickson polynomial problem, performing cross-checks using modular arithmetic. This aligns with its #7 global ranking in math on Arena.
Non-Math Reasoning
The model struggled with non-math reasoning, failing to properly check a timeline in a narrative logic puzzle, leading to an incorrect conclusion.
Conclusion
Qwen 3.7 Max is a strong model for hard problems but falls short in creative and narrative tasks. Official API pricing and full release are expected after the Alibaba Cloud Summit on May 20.