When running via Docker Compose, data loads automatically from the live database.
To load data manually, export from Open WebUI by asking the AI: "export my AI impact data as JSON",
then upload the resulting data.json file below — or load the demo data to explore.
ℹ️ You are viewing demo data. To see your real usage, export data from Open WebUI and upload it above.
Total Queries
—
AI requests
Total Tokens
—
input + output
Energy Used
—
mWh
CO₂ Emitted
—
mg CO₂
Water Used
—
mL
Est. Cost
—
USD
Local AI Savings
—
vs. cloud equivalent
Daily CO₂ Emissions
Daily Water Usage
CO₂ by Model
Cost by Model
Savings by Local Model
Per-Model Breakdown
Model
Queries
Tokens
Energy
CO₂
Water
Cost
💰 Local AI Savings vs. Cloud API
Each local model is matched to its closest commercial API equivalent by capability tier.
Savings = what you would have paid if each query went to the cloud API instead.
Per Saad-Falcon et al. (2025):
local LMs (≤20 B params) answer 88.7% of real-world queries at frontier accuracy,
with Intelligence per Watt improving 5.3× from 2023–2025.
Local Model
Cloud Equivalent
API Rate (in / out per 1k)
Queries
Input Tokens
Output Tokens
You Saved
Recent Queries
Time (UTC)
Model
Tokens
CO₂
Water
Cost
Scientific References
[1] Luccioni, A.S., Viguier, S. & Ligozat, A-L. (2023).
Power Hungry Processing: Watts Driving the Cost of AI Deployment?arXiv:2311.16863 /
DOI:10.1145/3627673.3679071.
[2] Li, P., Yang, J., Islam, M.A. & Ren, S. (2023).
Making AI Less "Thirsty": Uncovering and Addressing the Secret Water Footprint of AI Models.arXiv:2304.03271.
[3] US EPA. (2022). eGRID 2022 — Emissions & Generation Resource Integrated Database.epa.gov/egrid.
[4] Saad-Falcon, J., Narayan, A. et al. (2025).
Intelligence per Watt: Measuring Intelligence Efficiency of Local AI.arXiv:2511.07885.
Local LMs (≤20B params) answer 88.7% of real-world queries at frontier accuracy; IPW improved 5.3× (2023–2025).