Sustainable AI Group Released Energy Consumption Dashboard
The platform quantifies the energy demands of proprietary AI models, highlighting significant disparities in power efficiency.
Updated on Sept. 30, 2026 in Artificial Intelligence
Live Poll
Would you choose an AI model based on its energy efficiency rating?
The Sustainable AI Group has launched a dashboard that estimates energy consumption for closed AI models, revealing that agentic tasks can be significantly more power-intensive than standard chat interactions. This research-based project offers the first granular look at power requirements for specific proprietary systems like Kimi 3, DeepSeek v3, and Claude Fable 5.
Why it matters
Because model providers currently do not publish detailed energy data, this dashboard enables organizations to make informed decisions when routing workloads to the most efficient models. The release addresses a critical transparency gap in the AI industry as companies grapple with the environmental impact of scaling large-scale compute.
Claude Fable 5 consumes 76.2 Watt hours per 1-2 hour agentic session, an amount of energy equivalent to 183 seconds of air-fryer use. Notably, Fable 5 uses 30 times more energy per token than GPT-5 nano, and larger models generally require four times more power than their lighter variants.
The players
Sustainable AI Group
A research firm focused on quantifying the environmental footprint and energy consumption of large-scale artificial intelligence systems.
Etsy
An e-commerce company collaborating with researchers to refine methodologies for measuring the energy impact of AI workloads.
The details
The Sustainable AI Group calculates energy use by comparing proprietary models to open models of similar capability that have been subjected to experimental testing. The methodology estimates energy consumption on a per-token basis and subsequently maps these figures to estimated carbon emissions. Agentic sessions — automated sequences where an AI model executes multi-step tasks independently — were found to be 27 times more energy-intensive than standard conversational chat interactions.
Timeline
September 30, 2026: The Sustainable AI Group launched the dashboard and published initial energy data.
Early 2026: The Sustainable AI Group research firm was officially established.
The Tech Race
This release follows the framework established by the CLEER dataset to categorize the efficiency of proprietary AI systems. It creates a new competitive benchmark where energy footprint is as critical a metric as latency or token generation speed.
Organizations and developers can now use this dashboard to select more efficient AI models for their specific computational tasks, potentially reducing their operational energy costs. While the tool is currently available for comparative analysis, the inclusion of water consumption metrics is planned for a future release.
The takeaway
The study suggests that agentic workflows require significantly more power than traditional chat interfaces, a factor businesses should weigh when designing automated systems. Watch for the Sustainable AI Group's future updates, which aim to incorporate water consumption metrics into their existing efficiency dashboard.
Further reading
For broader context on the industry's environmental footprint, explore our coverage of Artificial Intelligence.
Live Poll
Would you choose an AI model based on its energy efficiency rating?
