Group 1 - The core issue facing tech giants is how to effectively integrate AI into end devices, with Honor proposing a dual approach focusing on AI models and underlying architecture to address power consumption, performance, and privacy challenges [1][2] - Honor's upcoming Magic8 series and MagicPad3 Pro will be the first to feature the fifth-generation Snapdragon 8 mobile platform, marking a significant step towards global market expansion [1][4] - The collaboration between Honor and Qualcomm aims to reshape AI productivity on devices, promoting a more inclusive, secure, and economical "Chinese model" [1][4] Group 2 - The relationship between chip manufacturers and terminal manufacturers is evolving from a simple supply-demand model to a more integrated partnership in the AI era [2][5] - Honor introduced a "high-efficiency end-side AI model solution" that includes two major technological breakthroughs: low-bit quantization technology, which reduces model storage space by 30%, increases inference speed by 15%, and decreases power consumption by 20% [2][3] - The second breakthrough is a new generation of vector retrieval technology that enhances search performance by 400%, allowing for rapid semantic understanding and matching of unstructured data [3][4] Group 3 - The combination of these technologies makes it feasible to build personal knowledge bases on mobile devices, showcasing the potential for releasing end-side AI productivity [4][6] - The competition in the mobile industry is shifting towards end-side AI, with a focus on user experience rather than just the presence of AI features [5][6] - Various companies, including Honor, Vivo, and Apple, are pushing towards an AI-centric future, indicating a significant shift in hardware, software, and service integration [6]
30亿参数模型已成各家旗舰机起跑线
Nan Fang Du Shi Bao·2025-09-28 23:15