Mistral AI reveals first comprehensive environmental audit of large language models

Mistral has released what it calls the first comprehensive environmental audit of a large language model, revealing the carbon emissions and water consumption of its “Large 2” AI model over 18 months of operation. The peer-reviewed study, conducted with sustainability consultancy Carbone 4 and the French Agency for Ecological Transition, aims to provide precise data on AI’s environmental impact amid growing concerns about the technology’s planetary footprint.

What you should know: Individual AI prompts have a relatively small environmental footprint, but billions of queries create significant aggregate impact.

• A single average prompt (generating 400 tokens of text) produces just 1.14 grams of CO₂ emissions and consumes 45 milliliters of water.
• Over 18 months, Mistral’s model generated 20.4 kilotons of CO₂ emissions—equivalent to 4,500 cars operating for a year—and consumed 281,000 cubic meters of water, enough to fill 112 Olympic swimming pools.
• The vast majority of environmental impact (85.5% of CO₂ emissions and 91% of water consumption) occurred during model training and inference rather than data center construction.

The big picture: The environmental cost of AI queries is comparable to other common internet activities, challenging assumptions about AI’s uniquely destructive impact.

• One Mistral prompt generates the same CO₂ emissions as watching 10 seconds of streaming video in the US or 55 seconds in France.
• A single query equals four to 27 seconds of a Zoom call, according to Mozilla Foundation data.
• Writing a 10-minute email read by 100 recipients produces emissions equivalent to 22.8 Mistral prompts.

Why this matters: The study provides rare transparency in an industry where environmental impact data is typically opaque, potentially setting a precedent for AI accountability.

• Despite widespread concerns about AI’s environmental effects, “it’s surprisingly hard to find precise, reliable data on the CO₂ emissions and water use for many major large language models.”
• Mistral’s numbers align with previous academic estimates, including a UC Riverside study showing 17ml of water per GPT-3 prompt and Nature research indicating 2.2g of CO₂ per ChatGPT query.

What they’re saying: Industry experts view the audit as an important first step toward standardized environmental reporting.

• “A great first step in terms of environmental impact assessment of AI models,” said Sasha Luccioni, Hugging Face’s AI & Climate Lead, though she noted the report lacks some methodological details.
• Mistral argues such transparency “could enable the creation of a scoring system, helping buyers and users identify the least carbon-, water- and material-intensive models.”

Key limitations: Mistral acknowledges its data represents “a first approximation” of the model’s total environmental impact, with important estimates used for GPU lifecycle calculations and missing details on total energy consumption versus emissions.