AI/LLM · Multi-Cloud
Mistral AI
Mistral runs a multi-cloud architecture where its hosted La Plateforme and Studio offerings sit atop partner clouds (Google Cloud, AWS, Azure, SAP, IBM, Snowflake, NVIDIA, Outscale), while open-weight models are also deployable on‑prem via standard inference containers.[10][11] The production stack is therefore split between first‑party EU‑centric hosting for regulated workloads and distribution of models through major cloud marketplaces and managed services.
Foundation proximity: Native (own / self-hosted weights)
Foundational models & how they consume them
PrimaryMistral Large (frontier reasoning model family, including latest versions)· Mistral AI first-party API (La Plateforme / Studio) · Direct API
SecondaryMistral Small / Medium / Devstral (production-efficient and developer-tuned models, including Devstral 2 123B)· Mistral AI first-party API (La Plateforme / Studio) · Direct API
TertiaryCodestral (code-generation family, including Codestral 25.05)· Mistral AI first-party API (La Plateforme / Studio) · Direct API
TertiaryPixtral / Pixtral Large (vision-language models)· Mistral AI first-party API (La Plateforme / Studio) · Direct API
TertiaryMistral 7B (open-weight)· self-hosted / open-weight distribution via standard Mistral inference container and partner clouds · Native (own / self-hosted weights)
TertiaryMixtral 8x7B (open-weight Mixture-of-Experts)· self-hosted / open-weight distribution via standard Mistral inference container and partner clouds · Native (own / self-hosted weights)
TertiaryMistral Small 3.1 (open-weight, on-prem friendly)· self-hosted via Mistral standard inference container and cloud marketplaces · Native (own / self-hosted weights)
TertiaryDevstral 2 123B· AWS Bedrock (managed foundation model API) · Cloud-hosted (Bedrock/Vertex/Azure)
TertiaryMistral open-weight models (e.g., Mistral 7B, Mixtral 8x7B) on Bedrock· AWS Bedrock · Cloud-hosted (Bedrock/Vertex/Azure)
TertiaryMedium 3· Microsoft Azure AI Foundry model catalog · Cloud-hosted (Bedrock/Vertex/Azure)
TertiaryCodestral 25.05· Microsoft Azure AI Foundry model catalog · Cloud-hosted (Bedrock/Vertex/Azure)
TertiaryMistral open weights (European-licensed)· Azure AI model catalog (publisher: Mistral AI) · Cloud-hosted (Bedrock/Vertex/Azure)
Cloud & datastore
Multi-Cloudobject storage for model artifacts, datasets, and logs (e.g., S3-compatible across partner clouds)[10]vector databases used with Mistral embeddings and open-weight models (e.g., Qdrant, Pinecone, pgvector; typical for La Plateforme and self-hosting scenarios)[11]relational databases for accounts, billing, and orchestration metadata (likely PostgreSQL/MySQL, inferred from typical SaaS/API patterns)[10][11]Compliance
SOC 2 Type II (claims to comply with SOC 2 Type II framework, with reports available via Trust/Help Center)[3][4][7][12][15]ISO 27001 (claims to comply with ISO 27001 framework)[3][4][7][12][15]ISO 27701 (claims to comply with ISO 27701 framework)[3][4][7][12][15]GDPR (EU-incorporated company with GDPR-focused data processing and SCCs; EU data residency by default for La Plateforme)[3][4][6][9][11]Similar companies
AI21 LabsAleph AlphaAnthropicCohereContextual AIHugging FaceOthers that build on Mistral
BNP ParibasDoctolibFireworks AIHelsingOrangeSAPExplore
FAQ
Does Mistral AI use a foundational AI model?
Mistral AI uses Mistral Large (frontier reasoning model family, including latest versions), Mistral Small / Medium / Devstral (production-efficient and developer-tuned models, including Devstral 2 123B), Codestral (code-generation family, including Codestral 25.05), Pixtral / Pixtral Large (vision-language models), Mistral 7B (open-weight), Mixtral 8x7B (open-weight Mixture-of-Experts), Mistral Small 3.1 (open-weight, on-prem friendly), Devstral 2 123B, Mistral open-weight models (e.g., Mistral 7B, Mixtral 8x7B) on Bedrock, Medium 3, Codestral 25.05, Mistral open weights (European-licensed) (Native (own / self-hosted weights)).
What cloud does Mistral AI run on?
Mistral AI runs on Multi-Cloud.
What database does Mistral AI use?
Mistral AI uses object storage for model artifacts, datasets, and logs (e.g., S3-compatible across partner clouds)[10], vector databases used with Mistral embeddings and open-weight models (e.g., Qdrant, Pinecone, pgvector; typical for La Plateforme and self-hosting scenarios)[11], relational databases for accounts, billing, and orchestration metadata (likely PostgreSQL/MySQL, inferred from typical SaaS/API patterns)[10][11].
How close does Mistral AI run to the foundational model?
Mistral AI is Native (own / self-hosted weights) — close to the metal, so cheaper tokens and more control over outputs.
Sources
mistral.ai ↗Also on the map
← Full orbit map (260 companies) · Score your own stack →
Architecture inferred from public sources · confidence medium · verify before betting on a detail.