Back to products
Mistral Small 3

Mistral Small 3

High performance in a 24b open-source model

Overview

What it is

- We’re committed to empower the AI community with open technology. Our open models sets the bar for efficiency, and are available for free, with fully permissive license. - Our optimized commercial models are designed for performance and are available via our flexible deployment options.

Intent

I need it when

Run agentic AI workflows for document analysis and data extraction

Mistral Small 3 supports multimodal and agentic capabilities, allowing it to process documents, extract structured data, and orchestrate multi-step tasks through Mistral Studio, with complementary OCR and specialized models available for document-heavy workflows.

Deploy AI models with privacy and data sovereignty controls

Mistral Small 3 can be self-hosted on-premises, in private cloud, or via EU-hosted Mistral infrastructure, allowing organizations to maintain data residency and avoid third-party data processing, critical for regulated industries and privacy-conscious enterprises.

Build cost-effective AI applications with lightweight inference

Mistral Small 3 is positioned as a cost-sensitive model option ($0.1/M input, $0.3/M output tokens) that maintains multimodal and multilingual capabilities, enabling developers to deploy production AI at lower operational cost than larger models while retaining reasonable performance for many use cases.

Integrate open-source AI models into custom applications

Mistral Small 3 is released as open-weight under Apache 2.0 license (for research/individual use) or commercial Mistral license, enabling developers to download, fine-tune, and deploy the model within their own systems with full control and customization.

Drop

Not a fit when

  • User requires on-device inference without API calls or self-hosting capability
  • User needs guaranteed real-time latency under 100ms for production systems
  • User requires models optimized specifically for non-English languages as primary use case
  • User operates in highly regulated industries requiring model transparency beyond Mistral's documentation
  • User needs extensive fine-tuning support without enterprise contract
Commercials

Pricing

API pricing per million tokens (input and output); Vibe subscription plans (Free, Pro $14.99/mo, Team $24.99/user/mo, Enterprise custom) View pricing