Have you ever wondered how much electricity you use when you encourage or appreciate an AI model? He embraced face engineer Julian De La Vande, and he built the tools to get to the answer.
AI models consume energy every time they run. They run on GPUs and specialized chips that require a lot of power to run the associated computational workloads at scale. While it is not easy to identify the power consumption of models, there is widespread expectation that the increased use of AI technology will bring new heights to electricity needs over the next few years.
The demand for more power to fuel AI has led some companies to pursue environmentally friendly strategies. Tools like Delavande are turning their attention to this and are likely aiming to pause AI users.
“Even small amounts of energy savings can scale to millions of queries. Model selection or output length can lead to major environmental impacts,” Delavande and The The The Creators wrote in a statement.
Have you ever wondered how much energy is used every time you send a message to ChatGpt?
We have created a version of the chat UI that shows in real time how much energy a message consumes. Should all chatbots view this?
Details below 👇👇pic.twitter.com/tbl0pw51pw
– Delavande Julien (@julidelavande) April 22, 2025
Delavande’s Tool is designed to work with Chat UI, the open source frontend for models such as the Meta’s Llama 3.3 70b and Google’s Gemma 3. This tool estimates the energy consumption of models and messages sent in real time, and reports watt-hour or joule consumption. It also compares the energy usage of the model with the energy usage of common household appliances such as microwaves and LEDs.
According to this tool, when you ask a typical email write to llama 3.3 70b, it uses about 0.1841 watts. This will run microwaves for 0.12 seconds or use a toaster for 0.02 seconds.
It is worth remembering that the only estimation of the tool is – estimate. Deravande does not claim that they are very accurate. Still, they act as a reminder that everything (with chatbots included) has costs.
“With projects like AI Energy Score and more extensive research into AI’s energy footprint, we are seeking transparency in the open source community. One day, energy use can be as visible as food nutrition labels!” Deravande and his co-creator wrote.
