Green AI: the energy behind the answers

Artificial intelligence consumes energy, a lot of it. Where and how models run makes the difference between an AI that burdens the environment and one that burdens it as little as possible. Ours falls into the second category.

How it works

Every response from your bot is calculated by dedicated processors, hosted in European data centers. Those data centers:

• draw electricity only from certified renewable sources;
• are designed for efficiency: in the newest facilities, the PUE reaches 1.2. PUE measures how much total energy is needed for every unit of energy that actually reaches the servers: 1 would be perfection, the global average is much higher;
• have monitored and reduced emissions for years, with plants continuously maintained and renewed.

The choice that matters most: tailored models

Self-hosting gives us an environmental advantage that few notice: we use models sized for the task. Answering a question about your shop's opening hours doesn't require the colossal model needed to write a thesis: it requires a skilled model, well trained, and the right size. Less computation per response means less energy per response.

What you need to do

Nothing. It is the way the platform is built.

Limits

• Green doesn't mean zero impact: AI still consumes energy. It means choosing every link in the chain, from energy to models, to consume as little as possible.
• We do not publish third-party certifications here: environmental documentation for the data centers we use is available upon request.

From the app

Not applicable: no configuration needed, it is how the platform is built.