AI Energy Score

AI Energy Score is an initiative to establish comparable energy efficiency ratings for AI models, helping the industry make informed decisions about sustainability in AI development.

To account for AI’s diverse applications, it benchmarks models across 10 distinct tasks spanning multiple modalities, testing both open and proprietary models. These results are made available in a public leaderboard to track progress over time:

The outcomes of the AI Energy Score analysis can be shared through a uniform label. The label includes the model’s name, GPU energy score, task, scoring date, benchmarking hardware, visual star rating, and link to the leaderboard for verification purposes. Following submission and leaderboard update, the label generator tool can be used to easily create and download a label:

For more information about the project, check out the AI Energy Score landing page and accompanying documentation

Project co-leads: Boris Gamazaychikov and Sasha Luccioni