
Mistral AI, a company specializing in artificial intelligence, announced the launch of its new "Mistral Medium 3" model, asserting it offers advanced performance at a low price compared to competing models.
According to the company, the new model achieves performance equivalent to or surpassing 90% of the results of Anthropic's "Claude Sonnet 3.7" model, despite its usage cost being significantly lower.
This version is priced at $0.40 per million input tokens and $2 per million output tokens when used via the company's programming interface.
Based on standard performance evaluations, "Mistral Medium 3" outperforms a range of prominent open models, such as Meta's "Llama 4 Maverick" and Cohere's "Command A".

The company also demonstrated its cost-effectiveness when self-hosted compared to models like DeepSeek v3.
Benefits of Mistral Medium 3
Mistral developed the model to operate efficiently in diverse environments, including cloud systems or standalone systems utilizing four or more GPUs.
It was specifically designed to support tasks related to programming, science, technology, engineering, and mathematics (STEM), providing high-accuracy multimodal understanding.
Companies from the finance, energy, and healthcare sectors have already begun using it to improve customer service, automate processes, and analyze complex data.
The company indicates that the model allows for flexible integration, whether through fine-tuning via its programming interface or through custom deployment within companies' private infrastructure.
Launch of Le Chat Enterprise
Additionally, Mistral announced the launch of "Le Chat Enterprise," an enterprise-focused version of its smart assistant.
This new service includes tools for building intelligent agents and integrates easily with productivity applications like Gmail, Google Drive, SharePoint, and others.
Furthermore, it enables companies to create custom assistants without the need for programming, and supports secure integration with internal data under strict privacy controls.
"Le Chat Enterprise" is available today via the Google Cloud Marketplace.
It will also be rolled out soon on other platforms such as Azure AI and AWS Bedrock, alongside anticipated support for the MCP standard, which enhances the connection of AI assistants to internal systems.