Mistral AI's New Model: Leading Performance & Value

Mistral Launches Efficient New AI Model: Mistral Medium 3

A new AI model, Mistral Medium 3, has been released by the French AI startup, Mistral. This model is specifically engineered for optimal efficiency while maintaining a high level of performance.

Pricing and Performance

The model is accessible through Mistral’s API at a cost of $0.40 per million input tokens and $2 per million output tokens. Mistral asserts that its performance is “at or above” 90% of that offered by Anthropic’s more expensive Claude Sonnet 3.7 model across various benchmarks.

Furthermore, Mistral Medium 3 demonstrates superior results when compared to recently released open-source models. This includes models like Meta’s Llama 4 Maverick and Cohere’s Command A, based on standard AI performance assessments.

Understanding Tokens

It’s important to understand that tokens represent the fundamental units of data processed by these models. Approximately 750,000 words constitute one million tokens – a quantity exceeding the word count of “War and Peace” by roughly 163,000 words.

According to a blog post shared with TechCrunch, Mistral Medium 3 is designed for deployment across diverse cloud environments. This includes self-hosted setups utilizing four or more GPUs.

Company Background and Funding

Founded in 2023, Mistral operates as a frontier model lab. Their objective is the development of a comprehensive suite of AI-driven services.

These services encompass a chatbot platform, Le Chat, and dedicated mobile applications. The company has secured over €1.1 billion (approximately $1.24 billion) in funding from venture capital firms, including General Catalyst.

Notable clients of Mistral include prominent organizations such as BNP Paribas, AXA, and Mirakl.

Ideal Use Cases

Mistral highlights that Mistral Medium 3 is particularly well-suited for coding and STEM-related tasks. It also exhibits strong capabilities in multimodal understanding.

Beta testing with clients in the financial services, energy, and healthcare sectors has focused on applications like customer service, automation of workflows, and in-depth analysis of complex datasets.

Availability and Integration

Beyond Mistral’s API, which allows for enterprise-level fine-tuning, Mistral Medium 3 is now available on Amazon’s Sagemaker platform. Integration with Microsoft’s Azure AI Foundry and Google’s Vertex AI platforms is planned for the near future.

Recent and Upcoming Releases

The release of Mistral Medium 3 follows the launch of Mistral Small 3.1 in March. The company has also indicated that a significantly larger model is slated for release in the coming weeks.

Le Chat Enterprise Launch

Mistral also announced the general availability of Le Chat Enterprise, a chatbot service tailored for corporate use. This service provides tools such as an AI agent builder and integration with popular third-party applications like Gmail, Google Drive, and SharePoint.

Le Chat Enterprise will soon incorporate support for MCP, Anthropic’s standard for connecting AI assistants to existing systems and software. This move aligns with announcements from other major AI providers, including Google and OpenAI, regarding MCP adoption.

Note: A pricing error in the initial blog post received by TechCrunch has been corrected. We apologize for any inconvenience this may have caused.