3.0× less energy than Devstral 2.
Devstral 2 Environmental Impact
Open-weight coding model — laptop-runnable
One query, in context
Energy, carbon and water for a single use. Code completion (~500 tokens)
About 2.0× a Google search and 72% of a minute of laptop use.
25% of a Google search.
About 12× a drop of water and 12% of a teaspoon of water.
- Where it runsMistral AI (France) / self-hosted
- Grid carbon intensity50 g CO₂/kWh · 90% renewable
- Compared with a Google search2x more than a search (0.3 Wh)
- Water~1,667 queries to fill 1 litre
Where Devstral 2 sits
#127 of 199 by energy per query. Heavier than 62% of all models. Lit dots are the same kind of model; click any dot to open it.
Want something lighter?
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
- 129.9 Wh
- Carbon
- 6.5 g
- Water
- 130 mL
A month of 10 requests a day, 5 days a week. The energy is about 1.2× a kettle boil and 2% of a UK home's day.
Work out your own →- Energy
- 519.6 Wh
- Carbon
- 26.0 g
- Water
- 520 mL
A month of 40 requests a day, 5 days a week. The energy is about 4.7× a kettle boil and 7% of a UK home's day.
Work out your own →- Energy
- 2.60 kWh
- Carbon
- 129.9 g
- Water
- 2.60 L
A month of 200 requests a day, 5 days a week. The energy is about 24× a kettle boil and 35% of a UK home's day.
Work out your own →How does Devstral 2 compare?
Next to similar models. Bars use a log scale, so each step is a big jump.
Detailed breakdown
Energy
Devstral 2 is Mistral's open-source coding model, designed to run on laptops. At ~0.6 Wh per completion, it's more efficient than Codestral and optimised for local software engineering workflows.
Power source & carbon
Open-source and laptop-runnable. When run locally, it inherits the user's grid carbon intensity. Via Mistral API, benefits from France's clean grid.
Water use
At ~2.5 mL per completion via API. Zero water when run locally.
About Devstral 2
Devstral 2 is an open-source code assistant model from Mistral AI, released in December 15, 2025, that runs below the category average for energy consumption at 0.60 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. It open-weight coding model — laptop-runnable.
Devstral 2 benefits from running in France, one of the cleaner grid regions in our dataset at 50 gCO₂/kWh with 90% renewable energy. The same model running in a coal-heavy region would produce significantly more carbon per query.
These figures are estimates derived from hardware specifications and API benchmarks — Mistral AI has not published official energy data for Devstral 2. Actual consumption may vary significantly depending on batching, quantisation, and infrastructure optimisations that we cannot observe from outside.
Devstral 2 in context
Key insights
- Runs on a 90% renewable grid — among the cleanest AI inference locations
- Open-source weights — can be self-hosted on infrastructure you control
What does your Devstral 2 usage cost the planet?
Use the calculator to estimate your personal footprint based on how often you use Devstral 2.
Frequently asked questions
How much energy does Devstral 2 use per query?
Each Devstral 2 query consumes approximately 0.60 Wh of energy. This is 2x more than a traditional Google search (~0.3 Wh).
What is Devstral 2's carbon footprint?
Based on the carbon intensity of Mistral AI (France) / self-hosted, each query produces approximately 0.03 g of CO2. The grid in this region has a carbon intensity of 50 g CO2/kWh with 90% renewable energy.
How much water does Devstral 2 use?
Each query consumes approximately 0.60 mL of water, primarily used for cooling the data centers that process the request.
How does Devstral 2 compare to a Google search?
A Devstral 2 query uses 2x more than a Google search in terms of energy. A Google search uses approximately 0.3 Wh, while Devstral 2 uses 0.60 Wh.
Technical details
- Architecture
- Transformer (code-specialised)
- Context window
- 128,000 tokens
- Release date
- 2025-12-15
- Open source
- Yes
- Training data cutoff
- 2025-12