Alibaba's latest AI model ran autonomously for 35 hours to optimize code for its own custom chip
Alibaba's Qwen team unveiled Qwen3.7-Max, a proprietary model engineered for extended autonomous reasoning tasks that ran uninterrupted for 35 hours optimizing code for Alibaba's custom silicon. The model matches Claude Opus 4.6 on standard benchmarks while outperforming Chinese competitors DeepSeek V4 Pro and Kimi K2.6, signaling Alibaba's push into long-horizon agent capabilities and vertical integration of AI infrastructure. The robotics demonstration underscores ambitions beyond language tasks, positioning Qwen as a contender in the emerging market for models that can sustain complex multi-step problem solving without human intervention.
Modelwire context
Analyst takeThe 35-hour autonomous run is notable less as a benchmark number and more as a signal that Alibaba is deliberately targeting the infrastructure layer, optimizing its own chip with its own model, which collapses the boundary between AI vendor and hardware customer in a way most Western labs have not yet done at this level of vertical closure.
This is largely disconnected from recent activity in our archive, as we have no prior coverage to anchor it to. It belongs, however, to a broader pattern visible across the industry: Chinese labs moving from pure model competition toward full-stack control, where the model, the chip, and the deployment environment are co-developed. That strategy mirrors what NVIDIA fears most from hyperscalers generally, and it puts Alibaba in a different competitive frame than simply chasing Claude or GPT on leaderboards.
Watch whether Alibaba publishes reproducible details on the chip optimization task, specifically the before-and-after performance metrics on its custom silicon. If those numbers surface with third-party verification within the next two quarters, the vertical integration claim is substantive; if they don't, this reads as positioning ahead of a product announcement.
This analysis is generated by Modelwire’s editorial layer from our archive and the summary above. It is not a substitute for the original reporting. How we write it.
MentionsAlibaba · Qwen · Qwen3.7-Max · Claude Opus 4.6 · DeepSeek V4 Pro · Kimi K2.6
Modelwire Editorial
This synthesis and analysis was prepared by the Modelwire editorial team. We use advanced language models to read, ground, and connect the day’s most significant AI developments, providing original strategic context that helps practitioners and leaders stay ahead of the frontier.
Modelwire summarizes, we don’t republish. The full content lives on the-decoder.com. If you’re a publisher and want a different summarization policy for your work, see our takedown page.