Mistral AI Launches Free API, Cuts Prices, and Enhances Vision Capabilities

Mistral AI is making significant strides in its mission to democratize access to cutting-edge AI. Today, the company announced several updates that will benefit both individual developers and enterprise users. These include a free tier on their serverless platform, reduced pricing across all models, an enhanced version of Mistral Small, and free vision capabilities on le Chat.

Free Tier on la Plateforme

La Plateforme, Mistral AI’s serverless platform for tuning and building with Mistral models as API endpoints, now offers a free tier. This is a game-changer for developers looking to experiment, evaluate, and prototype without any upfront costs. Key features of the free tier include:

• Seamless Upgrade: Users can easily transition their endpoints to a commercial tier.
• Data Isolation: Full data isolation with a zero-retention option available at no cost.
• Deployment Flexibility: Deploy models on Mistral’s cloud partners (Azure, AWS, GCP) or your own infrastructure.

This free tier allows developers to get hands-on experience with Mistral’s powerful models and explore their potential applications in a low-risk environment.

Reduced Prices Across the Board

Mistral AI has optimized its endpoints for speed and efficiency, leading to significant price reductions across all models. Here are the updated prices:

• Mistral Nemo:

  • $0.15 / M input tokens (previously $0.3)
  • $0.15 / M output tokens (previously $0.3)
  • Price drop: 50%

• Pixtral 12B:

  • $0.15 / M input tokens
  • $0.15 / M output tokens

• Mistral Small:

  • $0.2 / M input tokens (previously $1)
  • $0.6 / M output tokens (previously $3)
  • Price drop: 80%

• Codestral:

  • $0.2 / M input tokens (previously $1)
  • $0.6 / M output tokens (previously $3)
  • Price drop: 80%

• Mistral Large:
  • $2 / M input tokens (previously $3)
  • $6 / M output tokens (previously $9)
  • Price drop: 33%

These price reductions make Mistral Large 2 the most cost-efficient frontier model, while smaller models like Mistral Small and Codestral become extremely affordable. The updated pricing will also be reflected on cloud platform partner offerings such as Azure AI Studio, Amazon Bedrock, and Google Vertex AI.

Enhanced Mistral Small

Mistral AI is proud to introduce Mistral Small v24.09, an enterprise-grade update to their small model. This new version offers:

• 22 Billion Parameters: A mid-point between Mistral NeMo 12B and Mistral Large 2.
• Cost Efficiency: Highly cost-effective for use cases like translation, summarization, sentiment analysis, and more.
• Reliability: Suitable for enterprise environments where full-blown general-purpose models are not necessary.

Mistral Small v24.09 is available under the Mistral Research License, providing customers with a flexible, fast, and reliable option for a wide range of applications.

Free Vision Capabilities on le Chat

In addition to these updates, Mistral AI has introduced free vision capabilities on le Chat using Pixtral 12B. This integration allows developers to leverage advanced image processing and analysis without additional costs, expanding the scope of what can be achieved with Mistral’s models.

Conclusion

Mistral AI's latest updates are a clear demonstration of their commitment to making high-quality AI accessible to everyone. Whether you're a developer looking to experiment or an enterprise seeking cost-effective solutions, these changes offer significant value and flexibility.