GPT-4o Mini Environmental Impact

Energy Consumption

GPT-4o Mini consumes approximately 0.14 Wh per query — roughly one-third the energy of GPT-4o. As a distilled model optimized for cost-efficiency, it requires significantly fewer GPU computations per token, making it suitable for lightweight tasks where full GPT-4o power is unnecessary.

Power Source & Carbon

Runs on the same Microsoft Azure infrastructure as other OpenAI models (US East Virginia and Sweden Central). Its lower compute requirements mean shorter GPU engagement times per query, reducing the effective energy draw despite running on the same hardware.

Water Usage

At approximately 0.5 mL per query, GPT-4o Mini has a negligible water footprint — comparable to a single drop of water. Its efficiency makes it one of the most environmentally friendly commercial AI models available.

How much energy does GPT-4o Mini use per query?

Each GPT-4o Mini query consumes approximately 0.14 Wh of energy. This is about the same as a traditional Google search (~0.3 Wh).

What is GPT-4o Mini's carbon footprint?

Based on the carbon intensity of Azure US East / Sweden, each query produces approximately 0.06 g of CO2. The grid in this region has a carbon intensity of 450 g CO2/kWh with 25% renewable energy.

How much water does GPT-4o Mini use?

Each query consumes approximately 0.50 mL of water, primarily used for cooling the data centers that process the request.

How does GPT-4o Mini compare to a Google search?

A GPT-4o Mini query uses about the same as a Google search in terms of energy. A Google search uses approximately 0.3 Wh, while GPT-4o Mini uses 0.14 Wh.

Architecture

Transformer (decoder-only, distilled)

Context window

128,000 tokens

Release date

2024-07-18

Open source

No