Source URL: https://simonwillison.net/2025/Jul/22/mistral-environmental-standard/
Source: Simon Willison’s Weblog
Title: Our contribution to a global environmental standard for AI
Feedly Summary: Our contribution to a global environmental standard for AI
Mistral have released environmental impact numbers for their largest model, Mistral Large 2, in more detail than I have seen from any of the other large AI labs.
The methodology sounds robust:
[…] we have initiated the first comprehensive lifecycle analysis (LCA) of an AI model, in collaboration with Carbone 4, a leading consultancy in CSR and sustainability, and the French ecological transition agency (ADEME). To ensure robustness, this study was also peer-reviewed by Resilio and Hubblo, two consultancies specializing in environmental audits in the digital industry.
Their headline numbers:
the environmental footprint of training Mistral Large 2: as of January 2025, and after 18 months of usage, Large 2 generated the following impacts:
20,4 ktCO₂e,
281 000 m3 of water consumed,
and 660 kg Sb eq (standard unit for resource depletion).
the marginal impacts of inference, more precisely the use of our AI assistant Le Chat for a 400-token response – excluding users’ terminals:
1.14 gCO₂e,
45 mL of water,
and 0.16 mg of Sb eq.
They also published this breakdown of how the energy, water and resources were shared between different parts of the process:
It’s a little frustrating that “Model training & inference" are bundled in the same number (85.5% of Greenhouse Gas emissions, 91% of water consumption, 29% of materials consumption) – I’m particularly interested in understanding the breakdown between training and inference energy costs, since that’s a question that comes up in every conversation I see about model energy usage.
I’d really like to see these numbers presented in context – what does 20,4 ktCO₂e actually mean? I’m not environmentally sophisticated enough to attempt an estimate myself – I tried running it through o3 (at an unknown cost in terms of CO₂ for that query) which estimated ~100 London to New York flights with 350 passengers or around 5,100 US households for a year but I have little confidence in the credibility of those numbers.
Via @sophiamyang
Tags: environment, ai, generative-ai, llms, mistral, ai-ethics, ai-energy-usage
AI Summary and Description: Yes
Summary: The text discusses Mistral’s comprehensive lifecycle analysis (LCA) of their AI model, Mistral Large 2, which examines the environmental impacts of its training and inference processes. This analysis stands out as one of the most detailed assessments in the AI industry regarding ecological considerations.
Detailed Description: The text provides insights into Mistral’s contribution to establishing a global environmental standard for AI by releasing its environmental impact data for the Mistral Large 2 model. This is significant for professionals in the fields of AI and sustainability as it highlights the increasing importance of understanding the environmental implications of AI technologies.
– **Methodology**:
– Mistral collaborated with Carbone 4 and the French ecological transition agency (ADEME) to conduct the lifecycle analysis, ensuring a thorough evaluation which was also peer-reviewed.
– **Key Environmental Impact Numbers**:
– **Training Mistral Large 2**:
– 20.4 ktCO₂e (carbon dioxide equivalent)
– 281,000 m³ of water consumed
– 660 kg of Sb eq (standard unit for resource depletion)
– **Inference using AI assistant (Le Chat)**:
– 1.14 gCO₂e per 400-token response
– 45 mL of water
– 0.16 mg of Sb eq
– **Breakdown of Energy and Resource Utilization**:
– Approximately 85.5% of greenhouse gas emissions during model training and inference.
– 91% of water consumption linked to these processes.
– 29% of materials consumption associated with both training and inference.
– **Concerns Raised**:
– The author expresses frustration regarding the bundling of training and inference numbers, indicating a desire for clearer breakdowns to better understand the energy costs associated with each aspect.
– A request for contextualization of the environmental footprint (20.4 ktCO₂e) is made, comparing it to more relatable metrics such as the equivalent of flights or household energy use.
This analysis adds a layer of consideration for AI professionals focusing on ethical practices and sustainability within AI development. The insights underscore an essential emerging trend in AI, where environmental stewardship is becoming increasingly crucial, aligning with global efforts to minimize technology’s ecological footprint.