Rankings / Mistral AI
mistral-medium-2508 via api.mistral.ai
Usage share 0.01% · OpenRouter rankings API (daily token share, 2026-08-04)
Missing for Tier 2: external pre-deployment testing. The tier is computed from this checklist — satisfying these requirements moves the badge, automatically.
What happens to your data: training-on-customer-data defaults, retention windows, residency options, and the enterprise-versus-consumer terms gap. A legal property, not a capability — it survives every model generation.
EU residency and GDPR posture are genuine strengths; paid La Plateforme traffic is not used for training by default, but free-tier and some product-surface defaults are less clean, and documentation is thinner than the majors'.
Whether it changes without warning: versioning discipline, changelog quality, deprecation policy, and observed silent changes. The signal no one else tracks.
Dated version labels (e.g. -2508) and a deprecation table in docs; changelog quality is inconsistent.
Whether an attacker can make it misbehave — direct jailbreaks against the model's own policies and indirect prompt injection in agentic tool use. Graded to the weaker of the two, because an attacker takes the easier path.
Independent security indexes have consistently placed Mistral's hosted models in the lower tier for attack resistance; moderation is available as a separate API rather than built into model behavior.
No published agentic-injection hardening or agent red-teaming results at time of entry.
Whether you can see how it was built and tested: model cards, published safety evals, external pre-deployment testing, and disclosure of changes. The mechanism behind the tier ladder.
Model announcements and basic cards exist, but no external pre-deployment testing and sparse published safety evals. Mistral publicly disputes the framing of company-level safety indexes.
“Performance accuracy on all benchmarks were obtained through the same internal evaluation pipeline.”
Whether it is certified and compliant: SOC 2, ISO/IEC 42001, HIPAA eligibility, EU AI Act readiness, and audit availability.
SOC 2 attested; EU AI Act code-of-practice signatory with EU jurisdiction as a structural advantage; thinner certification stack than hyperscaler-hosted rivals.
No regional pinning — the provider chooses where data is processed.
EU-hosted La Plateforme (primary selling point); hyperscaler deployments via Azure and Bedrock follow those platforms' regions.
Paid La Plateforme traffic is not used for training by default, but free-tier and some product-surface defaults are less clean.
No tracked changes for this model. That is absence of detection, not evidence of stability — it may mean the model is unwatched, not that it is unchanging.
No dated score readings recorded for this model yet. Readings are only entered where multiple real, dated third-party values exist — never interpolated.
retrieved 2026-08-03 — 2026-08-05
No tracked changes yet for this model.
Compare this model: Mistral Medium 3.1 + open compare view →