Core Insights - JD Industrial has launched Joy Industrial, the first industrial-focused large model, aimed at enhancing cost efficiency, compliance, and supply assurance in industrial scenarios [1][6] - The model leverages JD's extensive data from over 10 million self-operated SKUs and 8 million active enterprise customers, integrating 30% of native data from its smart supply chain [3][5] - JD has developed a comprehensive large model matrix, including models of various sizes (3B, 10B, 81B, 750B), to meet diverse business needs [5] Group 1 - Joy Industrial includes AI agents for demand, operations, and customs, as well as product and integration experts for downstream users [1] - The model's training optimizes cost, efficiency, and user experience, achieving a resource requirement of only 1/16 of that of general large models [6][7] - JD's large model has already produced over 14,000 intelligent agents, addressing more than 18% of work content across various sectors [5][6] Group 2 - The model employs a mixture of experts (MoE) architecture to enhance inference efficiency, achieving an 8-fold increase in throughput compared to general models [7] - Future plans include a three-step strategy for AI collaboration, evolving from single-task AI employees to an AI organizational structure and ultimately an AI industrial ecosystem [8] - JD Industrial's extensive multi-modal data enables the construction of intelligent agents for complex supply chain coordination, addressing issues like data silos and management complexities [6][8]