Green Llama: A Tool for Monitoring Energy Consumption and Sustainability in Local LLMs

Anyan Huang, Philippe A. Henry, Yongcheng Huang, Yiming Chen.

Group 4.

Paper.Source code.

Large language models (LLMs) have revolutionized artificial intelligence by delivering unprecedented performance in various applications. However, their substantial computational demands have raised critical concerns regarding high energy consumption and environmental impact, especially during infer- ence on local machines. This paper introduces Green Llama, an innovative command-line tool designed to monitor and ana- lyze the energy usage of LLMs across CPU, GPU, and RAM components. By providing prompt-based estimation of power consumption and translating these metrics into carbon emissions based on regional energy profiles, Green Llama bridges the gap between performance benchmarking and sustainability assess- ment. With integrated features such as mid-session summaries, detailed reporting, and comprehensive benchmark testing, the tool empowers developers, researchers, and organizations to monitor energy consumption, reduce operational costs, and make environmentally responsible deployment decisions.