Energy numbers

What the number means

Every prompt has a cost. TreadLightlyAI shows it: each conversation displays the energy used to produce its responses, and the number grows as the conversation does.

For scale, a typical conversation uses a small fraction of a phone charge. Small is not zero. Seeing the cost is what lets you decide whether a prompt is worth it.

Before you send

The leaf indicator next to the input shows a relative estimate for your unsent prompt: S, M, L, or XL, based on message length and the selected model. Attachments and active tools count toward it.

Use it to right-size the prompt before it goes:

• Cut context the model does not need.
• Switch to a smaller model when the task is simple.

Measured vs estimated

There are two kinds of energy number, and in most cases we keep them clearly apart.

• Measured: when the provider reports energy use per request, we show that number as it arrives.
• Estimated: when a provider reports no per-request data, we estimate the energy ourselves and mark the value with a tilde (~). The Claude models work this way, and so do a few green ones.

The formula, rates, and sources behind the estimates are in the energy methodology. For the single-glance view of what we run on and where we compromise, see the transparency page.

Track your totals

Each conversation carries a leaf pill with its running energy total. Tap it for a breakdown of what used the energy, plus a real-world comparison for scale.

Your profile page totals energy across all your conversations, so your overall use is a number you can check, not a guess.