Build Agents and Understand Long Docs with Mistral Medium 3 and NVIDIA NIM

Developers building powerful multimodal applications now have a new state-of-the-art model designed for enterprise-scale performance with Mistral Medium 3.

Mistral Medium 3 combines high-performance, efficiency, and versatility in a compact deployment footprint. Designed for commercial and on-prem use cases, this dense model runs efficiently on NVIDIA Hopper GPUs, for enterprise developers seeking flexibility and control.

Smarter, longer, and multimodal

Medium 3 introduces multimodal support, enabling the model to accept both text and image inputs and return text outputs, making it suitable for a wider range of AI-driven applications, from document parsing to visual QA systems. Developers can now feed the model scanned reports, charts, or visual prompts and receive high-quality responses.

With a 128K context window, the model can reason over large documents, execute multi-step workflows, and retain long dialogue history—critical for applications like legal contract analysis, agentic planning, and multi-turn customer support.

It supports high-quality output across languages spanning over 45 global and regional languages, including Hindi, Vietnamese, Catalan, Arabic, Hebrew, and Japanese—making it ideal for localized AI experiences.

Screenshot of the Mistral Medium 3 language model running in the API catalog user interface, where the model understands natural language and image and generates responses based on user input prompt. — *Figure 1. An example of Mistral Medium 3 model generating responses from a user prompt*

Ready for the real world

Mistral Medium 3 is post-trained and aligned for instruction following, making it well-suited for chatbots, copilots, and agents. Developers can get started out-of-the-box, with reliable performance in:

Programming and debugging
Mathematical reasoning
Document and code summarization
Visual understanding
Function calling and agent workflows

With a private license, Mistral Medium 3 empowers organizations to integrate powerful AI capabilities securely, without sacrificing deployment control or cost efficiency.

NVIDIA NIM

To accelerate and simplify deployment, Mistral Medium 3 is available as an NVIDIA NIM, which simplifies scalable deployment across clouds and on-prem accelerated systems. NIM packages models like Mistral Medium 3 into ready-to-deploy containers, enabling developers to spin up high-performance endpoints in minutes rather than hours to days.