- Deployment
- on-prem
- Hq Country
- France
- Hq City
- Paris
- Ownership
- private-eu-owned
- Founded
- 2023
- Training Default
- never
- Residency Options
- on-premises, including air-gapped GPUs, customer VPC, dedicated Mistral Cloud with full data residency
- Gov Access Exposure
- none-eu
- Eu Ai Act Role
- gpai
- Certs
- ISO 27001, SOC 2 Type II
- DPA available
- yes
- Zero Retention Available
- Yes
- Self Host Option
- Yes
- Retention Default
- In the on-premises and VPC deployments the models run inside the customer's security perimeter, so there is no Mistral-side inference retention to publish. Customers can fine-tune or post-train the underlying models on private repositories and distil lightweight variants — the weights are in the customer's hands, not rented.
- Retention Exceptions
- The product was renamed during the period covered by this pass: mistral.ai/products/mistral-code now 404s and the offering is sold as Vibe for Code. The launch announcement and the coding solutions page both still describe the same three deployment shapes — on-prem, VPC, dedicated cloud — so the substance is unchanged.
- Subprocessors
- none in the inference path for an on-premises deployment — Mistral's public list at trust.mistral.ai/subprocessors covers the hosted products
- Subprocessor Count
- 0
- Transfer Mechanism
- intra-eu
- Eu Procurement Ready
- pass
- Eu Procurement Reason
- The strongest commercial coding assistant in the registry for a European buyer, and the only one where the vendor, the models and the deployment are all inside the Union at once. Codestral, Codestral Embed, Devstral and Mistral Medium run on the customer's own GPUs — Abanca keeps core banking code on-prem, Capgemini deploys for 1,500+ developers in regulated industries — and the customer can fine-tune on private repositories, which no closed copilot in this batch offers. It stops short of the open-weights rows because the stack is mixed rather than permissive: Devstral is Apache-2.0, Codestral ships under the Mistral Non-Production Licence, and Mistral Medium is proprietary, so exit is cheap for some of it and not for all of it. Worth knowing before signing: the product is built on the open-source Continue project, whose repository went read-only in 2026 — Mistral maintains the enterprise distribution, but the upstream is no longer moving.
- Training Claim Basis
- stated
- Training Confidence
- high
- Residency Claim Basis
- stated
- Residency Confidence
- high
- Retention Claim Basis
- stated
- Retention Confidence
- high
- Subprocessors Claim Basis
- stated
- Subprocessors Confidence
- high
- Residency Class
- customer-controlled
- Portability / Exit Path
- open-weights-restricted
- Subprocessor Jurisdiction
- structurally-none
- Underlying model providers
- Mistral AI (Codestral, Codestral Embed, Devstral, Mistral Medium)