Matches Claude 3.7 at 1/8th the cost. Just now, "European OpenAI" Mistral AI released a new multimodal model.
Mistral Medium 3 focuses on programming and multimodal understanding, balancing performance and cost.
According to the official announcement, in various benchmark tests, Mistral Medium 3's performance reaches or exceeds 90% of Claude Sonnet 3.7, with significantly reduced costs.
Claude 3.7 Sonnet is priced at $3 per million input tokens and $15 per million output tokens, while Mistral Medium 3 costs:
$0.4 per million input tokens and $2 per million output tokens.
Overall performance, whether compared to the leading open-source model Llama 4 Maverick or the enterprise-level model Cohere Command A, shows Mistral Medium 3 holding its own.
Currently, the Mistral Medium 3 API is available on Mistral La Plateforme and Amazon Sagemaker, and will soon be launched on IBM WatsonX, NVIDIA NIM, Azure AI Foundry, and Google Cloud Vertex.
Let's take a closer look at the benchmark tests.
Mistral Medium 3 is particularly strong in programming and STEM tasks.
The official statement says its performance