Pricing

State-of-the-art performance. Simplified enterprise deployments. Cost-efficient.

Thinking model excelling in domain-specific, transparent, and multilingual reasoning.

Top-tier reasoning for high-complexity tasks and sophisticated problems.

Enhanced model for advanced coding agents.

Introducing the world’s best document understanding API.

State-of-the-art transcription model

SOTA. Multimodal. Multilingual. Apache 2.0.

Thinking model excelling in domain-specific, transparent, and multilingual reasoning.

Lightweight, fast, and proficient in over 80 programming languages.

The best open-source model for coding agents.

State-of-the-art performance on speech and audio understanding.

Low-latency speech recognition for edge and devices.

Fine-tune Ministral 8B for classification tasks, like moderation, sentiment analysis, fraud detection, and more.

Fine-tune Ministral 3B for classification tasks, like moderation, sentiment analysis, fraud detection, and more.

A classifier service for text content moderation.

State-of-the-art embedding model for code.

State-of-the-art model for extracting representation of text extracts.

Tap into your enterprise data with integrations for Google Drive, SharePoint, OneDrive, Google Calendar, and Gmail.

Get improved, personalized answers by connecting le Chat to your data and organizing into complete knowledge bases.

Enhances AI with built-in tools for code execution, web search, image generation, persistent memory, and agentic orchestration.

Execute and interpret code snippets within the chat interface.

Enhance your work, research, and learning with web search, complete with citations for accurate and up-to-date information.

Generate images based on user prompts and preferences.

Access to news articles via integrated news provider verification for enhanced information retrieval.

Easily record and access API call data for debugging and continuous optimization.

Vision-capable large model with frontier reasoning capabilities.

Vision-capable small model.

State-of-the-art Mistral model trained specifically for code tasks.

Custom-trained model to serve specific geographies, markets, and customers.

A 7B transformer model, fast-deployed and easily customisable.

A 7B sparse Mixture-of-Experts (SMoE). Uses 12.9B active parameters out of 45B total.

Mixtral 8x22B is currently the most performant open model. A 22B sparse Mixture-of-Experts (SMoE). Uses only 39B active parameters out of 141B.

Powerful model for on-device use cases.

Most efficient edge model.