Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{207967,
        author = {SRIJIB SAMANTA},
        title = {GREEN ARTIFICIAL INTELLIGENCE: OPTIMISING THE ENERGY CONSUMPTION AND CARBON FOOTPRINT OF LARGE LANGUAGE MODELS},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {3},
        pages = {3477-3485},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=207967},
        abstract = {Large language models (LLMs) have created new capabilities in language generation, reasoning, coding and knowledge access, but their environmental cost extends across experimentation, training, deployment, data-centre overhead and hardware manufacture. This paper examines how Green Artificial Intelligence can reduce energy consumption and carbon emissions without treating model quality as the only measure of progress. A structured narrative review and lifecycle analysis are used to synthesise evidence on efficient architectures, sparsity, quantisation, distillation, retrieval-augmented generation, hardware-aware serving, workload scheduling and carbon accounting. The paper proposes a Green LLM Optimisation Framework built around five actions: measure the baseline, match model capacity to task difficulty, optimise the compute stack, shift flexible workloads to lower-carbon times and locations, and report quality–energy–carbon trade-offs. The analysis shows that a single universal “energy per prompt” value is misleading because energy depends on model architecture, precision, prompt and output length, batch size, accelerator utilisation, cooling overhead and electricity-grid intensity. Training receives public attention, yet repeated inference can dominate operational impact when services run at scale. Consequently, sustainable LLM deployment requires optimisation at the level of tokens, requests, models, servers and data centres. Green AI should therefore be understood not as a restriction on innovation but as a discipline for producing useful intelligence with the least defensible lifecycle burden. The paper concludes with a practical measurement protocol, governance recommendations and a research agenda for transparent, comparable and carbon-aware LLM systems.},
        keywords = {Green AI; large language models; energy efficiency; carbon footprint; sustainable computing; quantisation; carbon-aware scheduling; inference optimisation.},
        month = {August},
        }