Mistral AI SAS

AI Lab

Legal name: Mistral AI SAS · HQ: FR · Private · Founded 2023

French frontier AI lab. Published the first full lifecycle analysis of a large language model (Mistral Large 2, with Carbone 4 and ADEME) — a landmark in AI environmental transparency. Compute is hosted in France on a nuclear-dominated grid roughly six times cleaner than the US average.

Disclosure Quality

Score

45/100

Transparency Grade

T-3

partial disclosure

Grades run T-1 (most transparent) to T-5 (least transparent).

Red flags identified (2)

  • RF-01 · Total energy consumption not disclosed
  • RF-02 · A greenhouse-gas emissions scope not disclosed

See the disclosure audit below for details.

Fiscal year 2026. Score computed from OpenButterfly T1–T6 disclosure framework audit. Confidence tier: C (Estimated, Documented Method).

Environmental Metrics

Fiscal year 2026. Each metric is shown at its most recent available year; rows from a different year are marked.

Confidence tiers: A measured and verified · B company-reported · C documented estimate · D uncertain estimate · E modeled · ND not disclosed.

Total Energy Consumption

237 GWh

≈ the annual electricity use of 23,000 US homes

method: known_capacity_model

Source: OpenButterfly pipeline

C

Scope 1 Emissions

284 tCO₂e

method: scope1_ratio_model

Source: OpenButterfly pipeline

E

Scope 2 Emissions (market-based)

11 kt CO₂e

≈ the annual emissions of 2,314 passenger cars

method: grid_carbon_model

Source: OpenButterfly pipeline

D

Scope 3 Emissions

22 kt CO₂e

≈ the annual emissions of 4,720 passenger cars

method: scope3_ratio_model

Source: OpenButterfly pipeline

E

Water Withdrawal

296 ML

≈ 118 Olympic swimming pools of water

method: wue_model

Source: OpenButterfly pipeline

D

Property-Line Noise (estimated)

~57.2 dBA at property line

method: cooling_load_noise_model

Source: OpenButterfly pipeline

E

Disclosure Audit

Show 5 disclosure checks (T1–T6 framework)
T1.1

Total electricity consumption disclosed

No company-level energy reporting; disclosure is model-scoped, not operations-scoped.

RF-01 (Total energy consumption not disclosed)

Not Disclosed
T1.2

AI-specific energy consumption disclosed

First-ever full LLM lifecycle analysis (Mistral Large 2, with Carbone 4/ADEME): 20.4 ktCO2e and 281,000 m3 water disclosed for one model.

Partial
T2.1

Scope 1 GHG emissions disclosed

No corporate scope accounting published; LCA covers the model, not the company.

RF-02 (A greenhouse-gas emissions scope not disclosed)

Not Disclosed
T3.1

Total water withdrawal disclosed

Model-lifecycle water disclosed via LCA; company-level withdrawal not reported.

Partial
T6.1

GHG accounting methodology documented

Peer-reviewed LCA methodology with Carbone 4, ADEME, Resilio, Hubblo — a landmark in AI environmental transparency.

Fully Disclosed

Sources

← All entity profiles