Qwen3.7-Max Ran for 35 Hours on Unknown Hardware and Achieved a 10× Speedup
Alibaba's Qwen3.7-Max successfully ran for 35 hours on unfamiliar hardware, achieving significant improvements in kernel optimization. The model demonstrated its ability to adapt and refine its approach without human intervention, ultimately achieving a 10x speedup over the reference implementation. This performance surpasses other models in similar tasks, showcasing its unique capabilities in open-ended optimization problems.
- ▪Qwen3.7-Max completed a kernel optimization task on a hardware platform it had never encountered before.
- ▪The model made 1,158 tool calls and continuously improved its performance over 35 hours.
- ▪It achieved a 10x speedup, outperforming other models like GLM 5.1 and Kimi K2.6.
Opening excerpt (first ~120 words) tap to expand
HomeTechAlibaba's Qwen3.7-Max Ran Autonomously for 35 Hours on Unfamiliar Hardware. It Still... { "@context": "https://schema.org", "@type": "BreadcrumbList", "itemListElement": [{ "@type": "ListItem", "position": 1, "item": { "@type": "WebSite", "@id": "https://firethering.com/", "name": "Home" } },{ "@type": "ListItem", "position": 2, "item": { "@type": "WebPage", "@id": "https://firethering.com/tech/", "name": "Tech" } },{ "@type": "ListItem", "position": 3, "item": { "@type": "WebPage", "@id": "", "name": "Alibaba's Qwen3.7-Max Ran Autonomously for 35 Hours on Unfamiliar Hardware. It Still..." } } ] } Alibaba’s Qwen3.7-Max Ran Autonomously for 35 Hours on Unfamiliar Hardware. It Still Kept Getting Better.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at Firethering.