Evaluating and Comparing LLM Models for CO2 Footprint and Sustainable Green AI

Gaurav Sharma

Abstract: Large Language Models (LLMs) have significantly advanced artificial intelligence applications across various domains. However, these developments come at environmental cost in terms of carbon emissions, energy consumption, and water usage for cooling data centers. This paper evaluates and compares popular LLMs based on their sustainability and carbon footprint, analyzing key factors such as training and computing CO2 emissions, inference emissions, energy consumption, and water usage. We also discuss concerns and strategies for optimizing AI systems to reduce their environmental impact, emphasizing the need for responsible and sustainable AI development.

Keywords: Artificial Intelligence, Green AI, AI Co2 footprint, Sustainable AI, LLM (Large Language Models), Global Impact