AI BriefingMistral AI's Contribution to Establishing Global Environmental Standards for AI
Key point
Mistral AI conducted a Life Cycle Assessment (LCA) to quantify the environmental impact of LLMs and made the results public.
Details
Mistral AI conducted the industry's first LLM Life Cycle Assessment (LCA) to clearly understand the environmental footprint of AI technology. This research was carried out in collaboration with Carbone 4 and ADEME, and the robustness of the data was ensured through peer review by Resilio and Hubblo.
The analysis focused on three key categories: greenhouse gas (GHG) emissions, water consumption, and resource depletion. To increase transparency in the industry, Mistral AI disclosed the environmental impact generated during the training and inference of Mistral Large 2 as follows.
- Mistral Large 2 Training Impact (as of January 2025): 20.4 ktCO₂e emitted, 281,000 m³ of water consumed, 660 kg Sb eq of resource depletion
- Le Chat Inference Impact (based on a 400-token response): 1.14 gCO₂e emitted, 45 mL of water consumed, 0.16 mg Sb eq of resource depletion
These figures include not only the impact of GPU usage, carbon-intensive power grids, and water-stressed regions, but also upstream emissions generated during server manufacturing.
Mistral AI proposed three key indicators for future AI environmental management. These are the absolute impact of model training, the marginal impact of inference, and the ratio of total inference to the entire life cycle.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.