French AI startup Mistral is releasing the Mistral Medium 3, a new AI model focusing on efficiency without compromising performance.
Available for $0.40 per Mistral input token and $2 per million output tokens for $2 per million, the Mistral Medium 3 runs “up” 90% of human expensive Claude Sonnet 3.7 models, with a “board-wide benchmark.” It also outperforms recent open models, including Meta’s Llama 4 Maverick and Cohere’s Command A, and outperforms the popular AI performance rating.
A token is a raw bit of a data model that corresponds to 1 million tokens approximately 750,000 words (approximately 163,000 words longer than “war and peace.”
“Mistral Medium 3 can be deployed in the cloud, including self-hosted environments of four GPUs or more,” Mistral explained in a blog post sent to TechCrunch. “In pricing, this model beats cost leaders like Deepseek V3 in both API and self-deployment systems.”
Founded in 2023, Mistral is a frontier model lab and aims to build a variety of AI-powered services, including chatbot platforms, LE chats, and mobile apps. It is backed by VCS, including General Catalyst, and has raised more than 1.1 billion euros (approximately $1.24 billion) to date. Mistral’s customers include BNP Paribas, AXA and MIRAKL.
According to Mistral, Mistral Medium 3 is ideal for coding and STEM tasks and is excellent at understanding multimodal. The company says its financial services, energy and healthcare clients are beta-testing models of use cases, including customer service, workflow automation and analytics of complex datasets.
In addition to Mistral’s API, which allows enterprise customers to work with Mistral to fine-tune, Mistral Medium 3 will be available on Amazon’s Sagemaker platform starting Wednesday. It will soon be coming to other hosts, including Microsoft’s Azure AI Foundry and Google’s Vertex AI platform, the company added.
TechCrunch Events
Berkeley, California
|
June 5th
Book now
The launch of Mistral Medium 3 follows Mistral Small 3.1 in March. In a blog post, the company teased the release of a much larger model in the coming weeks.
Mistral on Wednesday also launched Le Chat Enterprise, a company-centric chatbot service that provides tools such as AI “agent” builders and integrates Mistral’s model with third-party services such as Gmail, Google Drive and SharePoint. Le Chat Enterprise was rolled out in a private preview earlier this year, but today it shows general availability.
Le Chat Enterprise will soon support MCP, the Anthropic standard for connecting AI assistants to systems and software where data resides. Other major AI model providers, including Google and Openai, announced earlier this year that they would adopt MCP.
Updated 7:48am Pacific: Fixed price error – Mistral’s blog post TechCrunch version had a typo. I regret the error