AI Energy Score launches star ratings for 166 models across 10 tasks
Hugging Face, Salesforce, Cohere and Carnegie Mellon publish an Energy Star-style leaderboard: GPU watt-hours per task, with five relative star bands. Open models are tested automatically; closed models through a sealed sandbox.
The AI Energy Score measures GPU energy for ten common tasks — text generation, summarisation, image generation and more — and ranks models within each task (and, for text generation, within size classes) into five 20% bands. Ratings are relative to the models evaluated at the time, so a five-star score means "among the most efficient tested", not an absolute threshold.
Because open-weight models can be benchmarked without the provider's cooperation, the leaderboard is the main independent check on companies that publish nothing themselves. Closed-model providers can submit through a secured sandbox; the index's transparency pillar rewards providers that do.
Sources
- 1.AI Energy Score — leaderboard and methodologyHugging Face, Salesforce, Cohere, Carnegie Mellon · Feb 10, 2025
- 2.Salesforce, partners launch energy ratings for AI modelsESG Today · Feb 11, 2025
