Today’s enterprises don’t just operate globally. They work in dozens of languages every day.
A legal team in Paris might need to review contracts in German, a sales team in Dubai may manage customer communications in Japanese, and support teams in Tokyo could be handling inquiries in Italian. While many AI models offer multilingual support, every organization has unique language requirements based on where they operate and the languages their teams actually use. Some businesses need strong performance across European languages, others prioritize Arabic or Hindi and many require robust support for Chinese, Japanese, or Korean.
That’s why we’re excited to announce that Box AI now supports models from Mistral AI, starting with Medium 3 and Small 3.1. With more options to evaluate, customers can choose the models that work best for their specific languages and use cases.
“At Mistral, we build models and AI solutions that blend top-tier performance with openness and accessibility for the enterprise,” said Michael Hoff, Global Head of Partners & Alliances at Mistral AI. “By integrating Mistral models into Box AI, customers can seamlessly query, summarize, and extract insights from their documents — all with enterprise-grade security, governance, and multilingual support. This collaboration makes it easier for organizations across the globe to operationalize advanced AI across their content workflows, delivering faster, more accurate insights where teams work.”
By integrating Mistral models into Box AI, customers can seamlessly query, summarize, and extract insights from their documents — all with enterprise-grade security, governance, and multilingual support.
Why Mistral AI models matter for global business
Enterprises today are more connected and linguistically diverse than ever before. As companies expand across continents, their content ecosystems become a blend of contracts, customer communications, and operational data in dozens of languages. Delivering consistent, high-quality work across that complexity requires AI that understands more than just words. It needs to grasp context, culture, and nuance.
Mistral AI's models are built precisely for this challenge. Known for their efficiency and advanced multilingual capabilities, Mistral AI’s Medium 3 and Small 3.1 models bring high performance and flexibility to global organizations using Box AI, empowering teams to work intelligently and securely across borders, languages, and content types.
Mistral Small 3.1 delivers impressive efficiency alongside robust multilingual support. With capabilities spanning dozens of languages, including English, French, German, Spanish, Italian, Portuguese, Dutch, Russian, Arabic, Hindi, Chinese, Japanese, Korean, Bengali, and Farsi, it can process extensive documents and conversations across language boundaries. The model also supports multimodal understanding, processing both text and images—valuable for teams working with diverse content types.
Mistral Medium 3 offers strong performance across complex reasoning, coding, and STEM tasks while maintaining cost-effectiveness. The model provides multilingual support across major European and Asian languages including English, French, German, Spanish, Chinese, Japanese, and Korean, and like Small 3.1, handles both text and image inputs.
Both models are trained to process multiple languages natively, rather than relying solely on translation between languages, which can help preserve context and nuance in multilingual workflows.
Powering multilingual use cases
With Mistral AI's multilingual models now integrated into Box AI, teams can work across languages without leaving Box’s secure environment. What was once a barrier between regional offices or global departments is now a bridge: Box AI can understand, summarize, and reason over content in dozens of languages, helping teams collaborate seamlessly across borders. The addition of Mistral AI's multilingual models to Box AI unlocks powerful new capabilities for our customers worldwide:
- Translation and localization: Sales teams can now quickly translate documents, contracts, and communications between languages while preserving context and nuance. This accelerates global collaboration and helps ensure nothing gets lost in translation.
- Cross-border contract analysis: Legal teams working on international agreements can analyze contracts in their original language, extracting key terms, obligations, and clauses without requiring manual translation—saving time and reducing the risk of misinterpretation.
- Multilingual customer support: Support teams can instantly summarize customer feedback, support tickets, and documentation in any language, enabling faster response times and better service for global customer bases.
- International compliance and governance: Security teams can automatically classify and govern content across multiple languages, ensuring compliance with regional regulations like GDPR while maintaining a unified content management approach.
- Global knowledge sharing: Research teams and knowledge workers can query content across language barriers, surfacing relevant insights from documents regardless of their original language—breaking down information silos that often emerge in multilingual organizations.
Built on Box’s trusted foundation
As with all Box AI capabilities, Mistral AI's models operate within Box’s robust security and compliance framework. Your content remains protected by Box’s enterprise-grade permissions, encryption, and governance controls. Users can only interact with content they’re authorized to access, and Box does not use customer data to train AI models. The integration also leverages Box’s existing infrastructure, so customers can access these powerful multilingual capabilities through the same familiar Box AI interface — whether working in Box’s web app, mobile apps, or through our extensive API ecosystem.
Get started
Mistral Medium 3 and Small 3.1 are now available now to Box customers via APIs and Box AI Studio. To get started with Mistral Medium 3 and Mistral Small 3.1, you can visit our linked documentation.
