Home/ Blog/ AI News/ Alibaba's latest AI model ran autonomously for 35 hours to o…

Alibaba's latest AI model ran autonomously for 35 hours to optimize code for its own custom chip

Alibaba's Qwen team releases Qwen3.7-Max, a proprietary model built for long-running autonomous agent tasks. It matches Claude Opus 4.6 on benchmarks and beats Chinese rivals like DeepSeek V4 Pro and Kimi K2.6. The team also demos the model steering a four-legged robot. The article Alibaba's latest AI model ran autonomously for 35 hours to optimize code for its own custom chip appeared first on The Decoder.

Alibaba's latest AI model ran autonomously for 35 hours to optimize code for its own custom chip

Alibaba’s Qwen team has unveiled its latest breakthrough, the Qwen3.7-Max, a proprietary AI model that recently demonstrated remarkable autonomy by operating independently for a staggering 35 hours. During this extensive period, the model dedicated itself to optimizing code specifically for Alibaba’s own custom-designed chip, showcasing a powerful fusion of AI and hardware development.

The newly released Qwen3.7-Max is engineered from the ground up for handling complex, long-running autonomous agent tasks, a capability brought vividly to life by its self-optimization feat. This extended, self-directed operation allowed the AI to iteratively refine and improve the performance of code intended for Alibaba’s proprietary hardware, hinting at a future where AI significantly accelerates chip design and optimization cycles. Beyond this, the Qwen team also offered a glimpse into the model's versatility, demonstrating its ability to steer a four-legged robot, illustrating its potential for tangible, real-world physical applications.

This development holds significant implications for the landscape of artificial intelligence and automation. The ability of an AI model to operate autonomously for such an extended duration, performing intricate optimization tasks, signals a major leap forward in AI agent capabilities. It moves beyond simple command-response systems towards truly self-sufficient problem-solving entities. Furthermore, Qwen3.7-Max's benchmark performance is noteworthy, matching the capabilities of leading models like Claude Opus 4.6. Domestically, it has also outperformed key Chinese rivals such as DeepSeek V4 Pro and Kimi K2.6, positioning Alibaba at the forefront of the competitive AI race, particularly within the autonomous agent domain.

Practically, Qwen3.7-Max could usher in a new era of efficiency across various sectors. For Alibaba itself, integrating this AI into its custom chip development process could drastically cut down design and testing times, leading to faster innovation cycles and more optimized hardware. Enterprises seeking advanced automation solutions will find the model’s long-running autonomous task capabilities particularly appealing, potentially transforming areas from complex software development to sophisticated industrial control systems. The competitive edge demonstrated against both international and domestic rivals also suggests a strengthening position for Alibaba in the global AI market, indicating increased pressure on other developers to match or exceed these sophisticated autonomous functionalities.

Alibaba's Qwen3.7-Max represents more than just another AI model; it embodies a strategic advancement towards deeply integrated AI autonomy in critical technological infrastructure. Its proven capacity for sustained self-optimization and its strong benchmark performance solidify its status as a leading contender in the next generation of AI agents, poised to redefine efficiency and innovation in both digital and physical domains.

Sources

Stay updated with daily AI news on Telegram. SYNC Hub automatically curates and delivers the most relevant AI & automation news directly in your feed.

Follow on Telegram →