Mistral AI

2026Mistral AIopen source

Voxtral Transcribe 2

v2.0
Agentic AI
Audio

Voxtral Transcribe 2 is a suite of next-generation speech-to-text models developed by Mistral AI, offering state-of-the-art transcription quality, diarization, and ultra-low latency. The suite includes Voxtral Mini Transcribe V2 for batch transcription and Voxtral Realtime for live applications, with the latter available under the Apache 2.0 license.

Voxtral Mini Transcribe V2: State-of-the-art transcription with speaker diarization, context biasing, and word-level timestamps in 13 languages.
Voxtral Realtime: Purpose-built for live transcription with latency configurable down to sub-200ms, enabling voice agents and real-time applications.
Best-in-class efficiency: Industry-leading accuracy at a fraction of the cost, with Voxtral Mini Transcribe V2 achieving the lowest word error rate, at the lowest price point.
version2.0
regionFrance
2026Mistral AIopen source

Leanstral 1.5

v1.5
featured
Formal Verification
Lean 4

Leanstral 1.5 is an open-source code agent model specifically designed for Lean 4, a proof assistant used for formal verification and mathematical research. It utilizes a Mixture-of-Experts (MoE) architecture to provide high-performance proof engineering and code verification capabilities while remaining cost-effective. It features 119 billion total parameters with 6.5 billion active per token; the model is engineered for efficiency and high performance, allowing it to tackle complex tasks like formalizing mathematical objects and verifying software specifications. It supports multimodal inputs, enabling users to process both text and images to generate rigorous, machine-verifiable proofs.

MoE architecture with 128 experts (6.5B parameters active per token)
256k token context length for long-range reasoning
Multimodal input support (text and image)
pricingFree
version1.5
2026Mistral AIopen source

Mistral Vibe CLI

v2.16.1
AI assistant
coding assistant

Mistral Vibe is an open-source command-line coding assistant developed by Mistral AI. It offers a conversational interface to your codebase, enabling natural language interactions for exploring, modifying, and managing projects directly from the terminal. Powered by Mistral's models, Vibe provides a suite of tools for file manipulation, code searching, version control, and command execution, all within a seamless CLI experience.

Interactive Chat: A conversational AI agent that understands your requests and breaks down complex tasks.
Built-in Toolset: A suite of tools for file manipulation, code searching, version control, and command execution, right from the chat prompt.
Project-Aware Context: Vibe automatically scans your project's file structure and Git status to provide relevant context to the agent, improving its understanding of your codebase.
pricingFree and open-source
version2.16.1
2025Mistral AIpaid

Mistral Vibe Code

vWeb
AI coding agent
open-source

Mistral Vibe Code is an AI coding agent designed to enhance software development workflows. It integrates seamlessly into terminals, IDEs, and background processes, providing developers with tools to write, test, and deploy code autonomously with a comprehensive codebase context.

Terminal-native coding agent
Multi-file orchestration
Smart references
versionWeb
regionFrance
2026Mistral AIopen source

Leanstral-2603

v1.0
AI Agents
Agentic AI

Leanstral-2603 is an open-source code agent developed by Mistral AI, designed specifically for the Lean 4 proof assistant. It combines multimodal capabilities with an efficient architecture, offering a performant and cost-effective alternative to existing closed-source solutions.

MoE: 128 experts, 4 active per token
Model Size: 119B parameters with 6.5B activated per token
Context Length: 256k tokens
pricingFree
version1.0
2026Mistral AIopen source

Mistral Forge

Agentic AI
Coding

Mistral Forge is a system designed for enterprises to build frontier-grade AI models grounded in their proprietary knowledge. It enables organizations to train models on internal data, ensuring alignment with unique operations and compliance requirements.

Pre-training on large internal datasets
Post-training for task-specific refinement
Reinforcement learning for policy alignment
regionFrance
2026Mistral AIpaid

Mistral Vibe 2.0

v2.0
Coding
AI Agents

Mistral Vibe 2.0 is a terminal-native coding agent powered by the Devstral 2 model family, designed to enhance code development efficiency. It offers features like custom subagents, multi-choice clarifications, slash-command skills, unified agent modes, and automatic updates.

Custom subagents
Multi-choice clarifications
Slash-command skills
version2.0
regionFrance
2026Mistral AIpaid

Mistral OCR 4

v4
CV
ML

Mistral OCR 4 is an advanced Optical Character Recognition (OCR) model designed to extract text and structured content from documents with high accuracy. It offers features like bounding boxes, block classification, and inline confidence scores, supporting 170 languages across 10 language groups. The model is available for self-hosted deployments and integrates seamlessly with Mistral's Search Toolkit for enterprise search and retrieval-augmented generation (RAG) pipelines.

Bounding boxes for text localization
Block classification (titles, tables, equations, signatures)
Inline confidence scores for extracted text
version4
regionFrance
2026Mistral AIopen source

Shieldstral

v1.0
Safety
Moderation

Shieldstral 1.0 is a 3B-parameter open-weights multimodal safety classifier that enables policy-adaptive content moderation. By framing moderation as a question-answering task, it allows developers to define safety policies in plain language at inference time without requiring model retraining.

Policy-adaptive moderation via natural language queries
Multimodal support for text, images, and text-image pairs
Efficient 3B parameter architecture optimized for single 16GB GPU deployment
pricingFree (Open Weights)
version1.0
2026Mistral AIopen source

Leanstral 119B A6B

v119B A6B
AI Agents
Agentic AI

Leanstral 119B A6B is an open-source code agent developed by Mistral AI, designed specifically for Lean 4, a proof assistant capable of expressing complex mathematical objects and software specifications. Built as part of the Mistral Small 4 family, it combines multimodal capabilities and an efficient architecture, making it both performant and cost-effective compared to existing closed-source alternatives.

MoE: 128 experts, 4 active per token
Model Size: 119B parameters with 6.5B activated per token
Context Length: 256k tokens
pricingFree
version119B A6B