Whisper Environmental Impact
OpenAI's open-source speech-to-text model
One query, in context
Energy, carbon and water for a single use. One minute of audio transcription
About 1.7× a Google search and 60% of a minute of laptop use.
About 1.8× a Google search and 10% of 10 m by car.
About 36× a drop of water and 36% of a teaspoon of water.
- Where it runsSelf-hosted (varies) / Azure
- Grid carbon intensity458 g CO₂/kWh · 32% renewable
- Compared with a Google search2x more than a search (0.3 Wh)
- Water~556 queries to fill 1 litre
Where Whisper sits
#117 of 199 by energy per query. Heavier than 57% of all models. Lit dots are the same kind of model; click any dot to open it.
Want something lighter?
Whisper is already among the lighter well-known options for this kind of work, so there isn’t a clearly lighter alternative to suggest.
What a month of use adds up to
Pick how much you use it. Based on one quick use at a time, five days a week.
- Energy
- 54.1 Wh
- Carbon
- 23.8 g
- Water
- 195 mL
A month of 5 uses a day, 5 days a week. The energy is about 4.5× a phone charge and 49% of a kettle boil.
Work out your own →- Energy
- 108.3 Wh
- Carbon
- 47.6 g
- Water
- 390 mL
A month of 10 uses a day, 5 days a week. The energy is about 9.0× a phone charge and 98% of a kettle boil.
Work out your own →- Energy
- 433.0 Wh
- Carbon
- 190.5 g
- Water
- 1.56 L
A month of 40 uses a day, 5 days a week. The energy is about 3.9× a kettle boil and 6% of a UK home's day.
Work out your own →How does Whisper compare?
Next to similar models. Bars use a log scale, so each step is a big jump.
Detailed breakdown
Energy
Whisper consumes approximately 0.5 Wh per minute of audio transcription. The largest model (1.55B parameters) is accurate but energy-intensive; the smaller 'tiny' (39M) and 'base' (74M) variants use a fraction of the energy at ~0.03-0.05 Wh/min. Whisper can run on consumer GPUs or even CPUs, making it one of the most widely deployed open-source AI models.
Power source & carbon
Fully open-source (MIT licence). Widely self-hosted on consumer hardware, cloud GPUs, and edge devices. When used via the OpenAI API, runs on Azure infrastructure.
Water use
At ~1.8 mL per minute of transcription (large model, cloud-hosted). Self-hosted on consumer hardware, water usage is zero.
About Whisper
Whisper is an open-source audio model from OpenAI, released in September 21, 2022, that runs well below the category average for energy consumption at 0.50 Wh per query. Because its weights are publicly available, it can be self-hosted on any infrastructure — meaning its carbon footprint depends entirely on where and how you choose to run it. At 1.55B parameters, it openai's open-source speech-to-text model.
These figures are estimates derived from hardware specifications and API benchmarks — OpenAI has not published official energy data for Whisper. Actual consumption may vary significantly depending on batching, quantisation, and infrastructure optimisations that we cannot observe from outside.
Whisper in context
Key insights
- Uses less than a third of the average energy for audio models
- Open-source weights — can be self-hosted on infrastructure you control
What does your Whisper usage cost the planet?
Use the calculator to estimate your personal footprint based on how often you use Whisper.
Frequently asked questions
How much energy does Whisper use per query?
Each Whisper query consumes approximately 0.50 Wh of energy. This is 2x more than a traditional Google search (~0.3 Wh).
What is Whisper's carbon footprint?
Based on the carbon intensity of Self-hosted (varies) / Azure, each query produces approximately 0.22 g of CO2. The grid in this region has a carbon intensity of 458 g CO2/kWh with 32% renewable energy.
How much water does Whisper use?
Each query consumes approximately 2 mL of water, primarily used for cooling the data centers that process the request.
How does Whisper compare to a Google search?
A Whisper query uses 2x more than a Google search in terms of energy. A Google search uses approximately 0.3 Wh, while Whisper uses 0.50 Wh.
Technical details
- Architecture
- Encoder-decoder Transformer (audio)
- Parameters
- 1.55B
- Release date
- 2022-09-21
- Open source
- Yes