ebb-ai

best-window finder

Plan a deferrable LLM workload

Pick a region, a deadline, and (optionally) a carbon budget. ebb-ai will choose the cleanest hour that still meets your deadline and show you the projected carbon cost.

Energy assumption: 1.50 gCO2e per task per 1000 g/kWh grid intensity — flat legacy estimate (1.5 Wh per task). Real dispatches (v0.10+) use per-model Wh/token coefficients at PUE 1.15.