The concept of Network-as-a-Service involves deploying and reconfiguring next-generation networks, in a flexible and dynamic manner, to always cater to the needs of the respective stakeholders. It presents a complex challenge to manage and orchestrate computational and telecommunication resources across the cloud-edge continuum, especially with the growing focus on cost efficiency and energy consumption. To address this complexity, several technology enablers are considered, but the recent advancements in Large Language Models research have inevitably brought Intent-Based Networking to the forefront. This paper explores the architecture and implementation of an intent-based automation framework that adheres to contemporary industry standards, while also considering sustainability. To achieve this, a translation pipeline is introduced, based on emerging multimodal Generative Artificial Intelligence models, which transforms a high-level description of desired network capabilities and supplementary deployment files into machine-consumable information digested by the network itself. Therefore, several state-of-the-art online and locally deployed models are compared. The ultimate motivation of this work is to validate the feasibility and accuracy of the proposed framework, promoting sustainability through minimal resource consumption and cost efficiency. Additionally, the framework ensures compatibility with modern orchestrators and next-generation Operational Support System that follow the same industry standards.