Mistral AI

European compute capacity for AI sovereignty by 2030.

Deploy AI models across diverse environments with full control.

Enables enterprises to train AI models on proprietary data.

Creating AI models optimized for Arabic language performance.

Developing AI models focused on cybersecurity and voice applications.

Platform supports third-party open models like Z.ai’s GLM-5.2.

Provides committed service levels for critical workloads.

Allows inference to run in Europe or the US.

3B multimodal safety classifier for content moderation.

8B model for single-camera autonomous robot navigation.

Open-source release of FLTEval for proof engineering evaluation.

Open-source AI model for formal verification and proof engineering.

Ensures uninterrupted AI tasks with workflow-integrated connectors.

Integrates connectors into Vibe Code for developer interfaces.

Provides root cause analysis for MCP connector issues.

Allows multiple account authentication for AI connectors.

Prevents impersonation in AI workloads with scoped API keys.

Enhances connector access control for secure AI integration.

OCR 4 integrated with Mistral's open-source search framework.

Advanced OCR model with multilingual support and bounding boxes.

Enhanced session controls and workflow automation in CLI.

Agent for long-running, multi-step work and coding tasks.

Agent for long-running, multi-step work and coding tasks.

Integrated AI stack for transforming industrial operations.

Framework for scaling neural operators across spatio-temporal problems.