Join waitlist
Back to products
Mistral 3

Mistral 3

A family of frontier open-source multimodal models

Website mistral.ai
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

Build and deploy custom AI agents for enterprise automation and long-horizon tasks

Mistral 3 enables users to create autonomous agents through Studio with end-to-end observability, custom workflows, and deployment portability across edge, cloud, and on-premises environments. Supports persistent memory, reusable skills, and 100+ enterprise connectors for knowledge integration.

Accelerate software development with AI-powered coding assistance

Vibe for Code provides terminal-native and IDE-integrated coding agents powered by Devstral, offering async code generation, legacy code translation, automated CI/CD reviews, and test authoring. Available in VS Code and JetBrains plugins with context-aware chat.

Train and fine-tune proprietary AI models on domain-specific data

Forge allows enterprises to turn proprietary knowledge into model intelligence through end-to-end model training, reinforcement learning, distillation, and synthetic data generation. Supports enterprise domain adaptation and evaluation lifecycle management.

Access cost-effective, high-performance language models for production applications

Mistral 3 provides open-weight models (Apache 2.0 licensed) and API access with transparent per-token pricing. Batch processing receives 50% discount. Models range from lightweight Mistral Small 4 ($0.1/$0.3 per M tokens) to flagship Mistral Medium 3.5 for reasoning and coding tasks.

Deploy frontier AI with full data privacy and control in regulated environments

Mistral 3 offers self-hosted deployment on virtual cloud, edge, or on-premises infrastructure with data remaining within organizational walls. EU-hosted cloud servers and enterprise-grade tooling support compliance requirements for financial services, public sector, and manufacturing.

Drop

Not a fit when

  • User requires guaranteed real-time response times or ultra-low latency for mission-critical systems without custom SLA negotiation
  • Organization needs proprietary closed-source models that cannot be self-hosted or customized
  • User requires extensive pre-built integrations with legacy enterprise systems not covered by Mistral's 100+ connectors
  • Project demands models optimized exclusively for non-English languages with no multilingual support
  • User needs on-device inference on resource-constrained devices without access to Mistral's Compute infrastructure
Commercials

Pricing

Freemium subscription with API token-based pricing. Free tier for basic use; Pro ($14.99/mo), Team ($24.99/user/mo), and Enterprise (custom) plans. API pricing per million tokens: Mistral Large 3 ($0.5 input, $1.5 output); Mistral Medium 3.5 ($1.5 input, $7.5 output); Mistral Small 4 ($0.1 input, $0.3 output). View pricing