Cleer Dashboard – How much energy⇄ does AI inference use? (cleerdash.sustainableaigroup.com)

The launch of the Cleer Dashboard introduces a novel tool for assessing the environmental impact of closed AI models through a method known as CLEER (Closed-model Latent Energy Estimation Range). This approach integrates peer-reviewed research, extensive empirical testing, and practical applications to help organizations accurately estimate the energy consumption associated with AI inference. By benchmarking open models and aligning proprietary models with their closest proxies, the dashboard translates complex per-token energy usage into tangible metrics for chat and agentic sessions, using public production datasets for context. This development is significant for the AI/ML community as it provides a standardized and transparent way to quantify the energy efficiency of different AI models, which is crucial given the rising concerns over the environmental impact of machine learning technologies. The dashboard captures not only the energy costs of model inference but also the broader environmental implications, including emissions from cooling, power distribution, and hardware construction. Users can adjust scenario parameters, such as the energy source and efficiency metrics, further tailoring the insights to specific operational contexts.