Alibaba unveils Qwen3.7-Max agent with 10x speed boost and 35-hour autonomy

Alibaba's Qwen team has unveiled Qwen3.7-Max, its most advanced AI agent model designed for autonomous, long-running tasks. The model excels in coding, debugging, office workflow automation, and can handle complex sequences spanning hundreds or thousands of steps. During internal tests on the company's new Zhenwu M890 chip platform, Qwen3.7-Max autonomously executed over 1,000 tool calls and iterative code modifications to optimize a key kernel. Alibaba claims this process improved inference speed by roughly 10x compared to the previous model generation.

Remarkably, the model sustained autonomous operation on complex tasks for up to 35 hours without human intervention. This positions Qwen3.7-Max as a powerful tool for enterprise automation, enabling agents that can manage entire workflows from code optimization to document processing. The model's upgraded reasoning and coding capabilities specifically target long-running agent workloads. API access is expected to roll out soon, giving developers and businesses a new option for building sophisticated AI agents. With its combination of extended autonomy, high-speed inference, and the dedicated Zhenwu M890 chip, Alibaba aims to compete directly with leading Western AI models in the agent space.