Mistral AI

Mistral AI develops efficient language models that balance performance and cost, from compact open-weight models to frontier-class API endpoints. Based in Paris, Mistral is the leading European AI company and a strong choice for applications with EU data residency requirements.

Key use cases include:

• Cost-efficient text generation and summarization
• Multilingual applications with strong European language support
• Code generation with Codestral, their specialized coding model
• Enterprise deployments requiring EU data sovereignty
• Self-hosted inference with open-weight models (Mistral 7B, Mixtral)
• Function calling and structured output extraction

Developers choose Mistral for the combination of open-source flexibility (self-host the open models) and premium API performance (Mistral Large rivals GPT-4-class models). The Mixtral mixture-of-experts architecture delivers high throughput at lower compute costs.

Mistral models integrate with all major frameworks including LangChain, LlamaIndex, and the Vercel AI SDK.

1. Sign up at console.mistral.ai and generate an API key.
2. Install the client:

   pip install mistralai

3. Set your API key:

   export MISTRAL_API_KEY='...'

4. Make your first call:

   from mistralai import Mistral
   client = Mistral()
   response = client.chat.complete(
       model='mistral-large-latest',
       messages=[{'role': 'user', 'content': 'Hello!'}]
   )

5. For open-weight models, download from Hugging Face or use hosted providers like Together AI or Groq.

Pricing: Mistral Large is ~$2/1M input tokens. Mistral Small is ~$0.20/1M input tokens. Open-weight models (Mistral 7B, Mixtral) are free to self-host. See mistral.ai pricing.