Mistral AI
Mistral AI
EU-hosted, regulation-aligned.
Impact Indicators
Job Displacement
No verified Mistral Large 3 or Mistral-specific realized job-loss data exists. Available peer-reviewed and working-paper evidence concerns generative AI adoption broadly, measuring early employment shifts across occupations rather than effects attributable to any single provider or model. Occupational effects reported in this research are not Mistral attribution. Mistral leadership commentary about AI and jobs is opinion, not measured employment evidence.
Privacy & Surveillance
Privacy evidence is product- and company-scoped, not specific to the `mistral-large-3` model. A security researcher disclosed a data-exfiltration technique affecting the Le Chat product; it is product-specific and there is no confirmed exploitation. A GDPR complaint has been filed against Mistral AI in France — status is a complaint filed only, with no investigation, finding, enforcement, settlement, or violation established. Mistral's own privacy policy describes its stated data-handling commitments.
Water Usage
There is no measured water figure for `mistral-large-3`, no measured Mistral-model-specific water figure, and no Le Chat-specific figure. Mistral has published a lifecycle environmental disclosure covering Mistral Large 2 — a different model version — whose water figure is modeled/estimated, not metered. No verified facility-level Mistral water measurement is available, and water withdrawal is not equivalent to water consumption. These estimates are not transferred to `mistral-large-3`.
Energy & Carbon
No independently verified energy or emissions figure exists for `mistral-large-3`, for any Mistral model specifically, or for Le Chat. Mistral's lifecycle disclosure covers Mistral Large 2, a different model version, and consists of company-disclosed modeled estimates. Broader data-centre electricity growth is documented only at industry level in national laboratory modeling. No verified Mistral facility electricity measurement exists, and GPU counts, FLOPs, capital expenditure, and megawatt capacity are not converted into measured energy use.
U.S. reporting requirements do not yet provide standardized, model-level disclosure for many AI human and environmental impacts. When U.S. evidence is incomplete, DoctorKnow uses verifiable EU regulatory disclosures, public-sector data, independent research, and reputable journalism to provide the strongest evidence currently available.
Evidence scope is identified as model-specific, provider-level, facility-level, regional/grid, or industry-level.
Scores use a 0–100 scale. Higher scores indicate stronger performance or lower concern; 100 is the best possible rating.
Reasoning
87
Coding
88
Writing
86
Analysis
85
Speed
86
Scores use a 0–100 scale, where 100 represents the strongest outcome and 0 the weakest.
What Mistral AI does
Mistral AI is built by Mistral AI and is positioned for code, writing and chat. Our readings score it strongest on code generation and multi-step reasoning.
Its weakest measured axis is data analysis (85/100), so plan around that when the work depends on it. Access is freemium, and the weights are closed, so you can only reach it through the vendor's own service.
Ethical stance
Transparency reads 77/100. Documentation, evaluations and model details are unusually open for the category.
Labor and job-displacement posture reads 66/100. There is partial disclosure on data work and workforce impact, with clear gaps.
Regulatory posture reads 91/100. Mistral AI's vendor engages constructively with binding AI rules.
Privacy and data handling
Privacy and data handling reads 84/100. Retention is short or optional, training on your prompts is off by default, and enterprise terms are clear.
Because the weights are closed, every prompt is processed on vendor infrastructure — treat regulated or confidential data accordingly.
Energy, carbon and water
Carbon footprint reads 81/100. Serving runs largely on low-carbon power, with credible reporting behind it.
Water and local impact reads 79/100. Cooling relies on closed-loop or air-side designs with limited draw on stressed watersheds.
Speed and efficiency also matter here: a 86/100 speed reading means fewer compute-seconds per answer for routine work.
Diagnostic notes
EU AI Act aligned, French low-carbon grid, GDPR-native data handling.
Readings are DoctorKnow.ai's own 0–100 assessments, compiled from vendor documentation and public reporting, and revised as new disclosures land. They are a comparison aid, not a certification.