Core Viewpoint - The collaboration between Paradigm Intelligence and Sunrise aims to significantly reduce AI inference costs to 1 cent per million tokens, facilitating the large-scale application of AI technology [1][9]. Group 1: Cost Reduction Initiative - The "1 cent per million tokens" inference cost plan is designed to address the high costs associated with large model inference, which currently ranges from 0.4 to 2 yuan per million input tokens and 1 to 4 yuan per million output tokens [3]. - The partnership leverages Sunrise's new generation inference GPU chip, the Yingwang S3, and Paradigm's PhanthyCloud cloud service to achieve a 90% reduction in unit token costs in typical scenarios [3][9]. Group 2: Industry Impact - This initiative is seen as a potential turning point for AI infrastructure development, moving the industry from "technical validation" to "scale application" [9]. - The collaboration is expected to enhance the practical application value of domestic computing power, supporting small and medium enterprises and government agencies in adopting intelligent solutions at low costs [7][9]. Group 3: Company Background - Sunrise is a Chinese AI computing chip company focused on developing large model inference GPUs, with a strategic financing of approximately 3 billion yuan completed in the past year [10]. - Paradigm Intelligence, founded in 2014, is a leading global AI technology company with a mission of "AI for Everyone," having implemented over 10,000 AI applications worldwide across various sectors [11].
范式智能&曦望|推出“百万Token一分钱”计划 重构大模型推理成本边界