The Generative EDGE AI Working Group, established since 2024 within the EDGE AI FOUNDATION, is dedicated to advancing the responsible adoption of generative artificial intelligence (AI) at the edge and on-premises. This paper provide a report of its operations outlines the group's mission to connect academia, industry, and the open-source community through collaborative research, educational initiatives, and community-driven activities. It presents a vision for deploying generative models, such as small language models (SLMs), multimodal AI, and agentic systems, for resource-constrained devices. In addition to detailing the group's objectives and deliverables, this work includes a comprehensive summary of the speeches delivered by leading industries and research entities during the third Forum of the GenAI on the Edge series. These summaries highlight both qualitative and quantitative insights shared by global experts, covering topics such as model optimization, hardware innovations, real-world applications, and emerging paradigms like agentic AI. By integrating these perspectives, the paper provides a holistic view of the current state and future potential of generative AI at the edge. This effort aims to establish a foundation for scalable, inclusive, and impactful innovation in generative intelligence, positioning the working group as a key reference point for the community driving the transformation from cloud centric to both on premises and edge centric edge AI
Expanding the horizons of generative edge AI: Mission, vision, and insights from industries
Submitted to TechRxiv, 26 July 2025
      
  Type:
        Report
      Date:
        2025-07-26
      Department:
        Communication systems
      Eurecom Ref:
        8438
      Copyright:
        © EURECOM. Personal use of this material is permitted. The definitive version of this paper was published in Submitted to TechRxiv, 26 July 2025 and is available at : 
      See also:
        
      PERMALINK : https://www.eurecom.fr/publication/8438
 
 
 
     
                       
                      