AI Energy Score unveiled for sustainable AI development

A new benchmarking tool, the AI Energy Score, has been launched to allow AI developers and users to evaluate the energy consumption of AI models.

Salesforce, in collaboration with Hugging Face, Cohere, and Carnegie Mellon University, announced the AI Energy Score, which serves as a platform for energy efficiency in AI models. This initiative introduces public ratings for 166 frequently-used AI models, enabling users to choose models that utilise less energy.

Salesforce will pioneer the disclosure of energy efficiency data for its proprietary models under this new framework. The initiative seeks to address the lack of transparency concerning the environmental impact of AI models by establishing a clear benchmark for sustainability, akin to the ENERGY STAR ratings for appliances and electronics.

Dr. Sasha Luccioni, AI & Climate Lead at Hugging Face, remarked, "The AI Energy Score represents a major milestone for sustainable AI. By creating a transparent rating system, we address a key blocker for reducing the environmental impact of AI. We're thrilled to launch this project and look forward to seeing wider adoption."

As a part of the initiative, the AI Energy Score was recognised by the French Government and the Paris Peace Forum. The score features a standardized energy ratings framework for AI models across various tasks, including text generation and image generation. A comprehensive leaderboard presents scores for common AI functions performed by 166 models, which includes Salesforce's SFR-Embedding, xLAM, and SF-TextBase.

Developers can submit their models, whether open or proprietary, to the benchmarking portal for evaluation and inclusion on the leaderboard. A new 1- to 5-star label will rate AI model energy use, guiding developers and users towards more sustainable choices.

Bruno Bonnell, General Secretary for Investment in France, commented, "The AI Energy Score exemplifies the missions of the AI Action Summit by tackling a pressing societal issue — AI transparency and sustainability — through bold innovation and global collaboration. We are proud to showcase this groundbreaking initiative at the Summit, as it stands as a powerful testament to the ingenuity of the global academic and industry ecosystem and a rallying call for collective action towards responsible AI development."

Ariane Thomas, Global Tech Director of Sustainability at L'OREAL Group, stated, "Transparency like that offered by the AI Energy Score is crucial. Over the past three years, we've collaborated to refine and obtain more granular elements for the key indicators like CPU and RAM data. This initiative sets an important milestone, and we hope more proprietary model providers join us in adopting such practices to drive accountability and sustainability across industries and supply chains. By openly sharing energy consumption data, our companies can collectively implement ecodesign practices and minimize the environmental footprint of AI technologies."

Suzanne DiBianca, EVP and Chief Impact Officer at Salesforce, added, "Reducing AI energy consumption lowers operational costs, optimizes infrastructure, and enhances long-term sustainability and profitability. We are proud to work with industry leaders to build a more transparent AI ecosystem."

Beyond benchmarks, Salesforce has addressed sustainability through the introduction of Agentforce, an agentic layer for deploying autonomous AI agents across various business functions, prioritising high performance while reducing environmental impact.

This initiative reflects a broader industry movement towards transparency and environmental responsibility in AI development, leveraging collaborations among leading technology and academic sectors.