Public Preview: Mistral Large 3 in Foundry

Mistral Large 3, a frontier-class open-weight model with Apache 2.0 licensing, is now available in Microsoft Foundry on Azure. This model delivers enterprise-grade reliability, long-context comprehension, and multimodal reasoning—making it ideal for production assistants, retrieval-augmented applications, agentic systems, and multimodal workflows.

Key Capabilities:

• Highly reliable instruction following
• Extended context support for long documents and multi-turn conversations
• Strong multimodal reasoning (text, images, diagrams)
• Stable, predictable performance for enterprise workloads
• Fully open and exportable for hybrid or on-premises deployment

Why it matters:
Mistral Large 3 sets a new benchmark for open models, combining frontier-level capability, reproducible results, and worldwide availability without regional restrictions. Organizations can fine-tune, customize, and deploy the model across cloud, edge, or sovereign environments with no attribution requirements.

Pricing:

• Input: $0.50 per 1M tokens
• Output: $1.50 per 1M tokens
• Deployment: Global Standard (West US 3)

Learn more.