Mistral Medium 3 Release Notes

Mistral AI has announced Mistral Medium 3, a new class of language model designed to provide frontier-level performance at a significantly lower cost and with greater deployment flexibility. This release aims to reduce the cost of enterprise AI integration by offering a model that competes with much larger, more expensive competitors while remaining deployable on relatively modest hardware.

Performance and Cost Efficiency

Mistral Medium 3 delivers frontier-class performance while being an order of magnitude less expensive than comparable models. It performs at or above 90% of Claude Sonnet 3.7 across benchmarks, with a pricing structure of $0.4 per million input tokens and $2 per million output tokens.

In terms of competitive positioning, Mistral Medium 3 surpasses leading open models such as Llama 4 Maverick and enterprise models such as Cohere Command A. It also beats cost leaders like DeepSeek v3 in both API and self-deployed system configurations.

Technical Capabilities and Professional Use Cases

Mistral Medium 3 is optimized for professional use cases, specifically excelling in coding, STEM tasks, and multimodal understanding. According to internal evaluation pipelines, the model's accuracy in these domains allows it to compete closely with larger, slower competitors.

Third-party human evaluations further support these findings, indicating that Mistral Medium 3 delivers superior performance in the coding domain compared to several of its larger competitors.

Enterprise Deployment and Customization

Mistral Medium 3 is designed for deep integration into enterprise systems, supporting various deployment modes to meet security and privacy requirements:

  • Deployment Options: The model supports hybrid, on-premises, or in-VPC deployments. It can be self-hosted on environments consisting of four GPUs or more.
  • Customization: Through Mistral's applied AI solutions, the model supports continuous pretraining, full fine-tuning, and blending into enterprise knowledge bases for domain-specific training and adaptive workflows.

Beta customers in the healthcare, energy, and financial services sectors are currently utilizing the model for complex dataset analysis, business process personalization, and enriching customer service with deep context.

Availability

Mistral Medium 3 is available immediately via Mistral La Plateforme and Amazon Sagemaker. Availability is expanding to include IBM WatsonX, NVIDIA NIM, Azure AI Foundry, and Google Cloud Vertex.

Regarding future releases, Mistral AI indicated that a new "large" model is expected to be released in the coming weeks.

Sources

Related

  • Dispatch
    Mistral 3 Release Notes: Mistral Large 3 and Ministral 3 Family

    Mistral AI has released Mistral 3, featuring the frontier-scale Mistral Large 3 MoE model and the edge-optimized Ministral 3 family (3B, 8B, 14B), all under the Apache 2.0 license.

  • Dispatch
    Mistral Code Release Notes

    Mistral AI has launched Mistral Code, a vertically integrated AI coding assistant for enterprises that combines frontier models with an in-IDE assistant and flexible deployment options.

  • Dispatch
    Mistral Small 3 Release Notes

    Mistral AI has released Mistral Small 3, a latency-optimized 24B-parameter model under the Apache 2.0 license that rivals the performance of models three times its size.

  • Dispatch
    Mistral AI Devstral Models Release

    Mistral AI has released Devstral Medium and Devstral Small 1.1, providing state-of-the-art agentic coding capabilities with a focus on generalization across different prompts and agentic scaffolds.

  • Dispatch
    Mistral Large 2 release notes / what's new

    Mistral AI has released Mistral Large 2, a 123B parameter model designed for high throughput single-node inference with competitive performance in coding, reasoning, and multilingual capabilities.