Alibaba has released what it calls its most advanced AI model yet: Qwen3.7-Max. Designed for complex, multi-step tasks, the model can autonomously execute over 1,000 tool calls in a single session, with inference speeds improved by roughly 10 times over previous versions.
The model was previewed on May 14 on the LM Arena leaderboard and formally announced at the Alibaba Cloud Summit in Hangzhou between May 19 and May 21. It is now available through the Alibaba Cloud Model Studio API, giving developers and businesses immediate access.
On key benchmarks, Qwen3.7-Max scored 92.4 on GPQA Diamond (graduate-level reasoning), 80.4 on SWE-Verified (real-world software engineering), and 91.6 on LiveCodeBench (coding problems beyond its training data). The model is compatible with OpenAI’s API format and works with agent frameworks including Claude Code and OpenClaw.
Alibaba says iterative kernel optimizations drove the performance gains, though no connection to digital assets or blockchain was mentioned. The model is positioned within its cloud and enterprise services ecosystem.