French AI startup Mistral has announced the launch of its newest artificial intelligence model, Mistral Medium 3, emphasizing strong efficiency gains without sacrificing superior performance.
According to the startup, Mistral Medium 3 delivers performance levels that are comparable to or exceed 90% of the capabilities of Anthropic’s significantly more expensive Claude Sonnet 3.7 model. Available immediately through Mistral’s API, the new model carries a cost of $0.40 per million input tokens and $20.80 per million output tokens. It has already surpassed recent open-source competitors such as Meta’s Llama 4 Maverick and Cohere’s Command A on popular AI benchmarking tests.
Mistral emphasizes that the Medium 3 model is engineered for affordability paired with high versatility, fitting seamlessly into cloud environments of various scales. The company specifically highlighted that the model can be effectively deployed in systems utilizing as few as four GPUs, thereby enhancing its appeal for businesses seeking manageable infrastructure costs.
Company representatives describe the new offering as particularly adept at coding tasks and STEM-related applications. It has also demonstrated notable strength in multimodal understanding. In testing phases, enterprises across sectors including financial services, healthcare, and energy utilized Mistral Medium 3 for applications such as customer service automation, data-intensive workflow management, and complex data analytics.
Mistral, founded in 2023, positions itself as a frontrunner in the development of frontier AI models, supported by substantial venture capital investment that exceeds €1.1 billion (approximately $1.24 billion). Its client roster includes prominent industry names such as BNP Paribas, AXA, and Mirakl, underscoring the practical enterprise demand for Mistral’s AI solutions.
Alongside direct access via Mistral’s API, businesses have the option to fine-tune the Medium 3 model for domain-specific applications. Moreover, the model’s availability is expanding, first becoming accessible on Amazon’s SageMaker platform this week, followed by future integrations into Microsoft’s Azure AI Foundry and Google’s Vertex AI platforms.
The launch of Mistral Medium 3 comes shortly after the release of Mistral Small 3.1 earlier this year. Additionally, Mistral reveals intentions to introduce an even larger and more powerful model in the coming weeks, signaling ongoing growth and competition in the AI modeling space.
