-
Models
It starts with frontier science.
State-of-the-art models with cutting edge capabilities, from cloud to edge.
Codestral
Devstral 2
OCR 3
Magistral Small
Voxtral TTS
Ministral 3 - 14B
Mistral Moderation
Codestral Embed
Leanstral
Classifier API model 3B
Mistral Medium 3
Model families.
Generalist models.
Open
Mistral Large 3
Open-weight, general-purpose, flagship multimodal and multilingual model.
Open
Mistral Medium 3.5
Our flagship model that merges instruction-following, reasoning, and coding into a single 128B dense model. Released as open weights, under a modified MIT license.
Open
Mistral Small 4
SOTA. Multimodal. Multilingual. Apache 2.0.
Open
Ministral 3 - 8B
Best-in-class frontier AI to the edge.
Specialist models.
Premier
OCR 3
The world's best document extraction and understanding model.
Open
Voxtral TTS
State-of-the-art text-to-speech generation and voice cloning.
Premier
Voxtral Mini Transcribe 2
State-of-the-art transcription model
Open
Voxtral Mini Transcribe Realtime
Open model fine-tuned and optimized for transcription.
Open
Voxtral Small
State-of-the-art performance on speech and audio understanding.
Open
Devstral 2
Open-weights agentic coding model for autonomous software engineering.
Premier
Codestral
Low-latency coding model optimized for high-frequency completion, fill-in-the-middle, and code generation tasks.
Premier
Codestral Embed
State-of-the-art embeddings for code and natural language queries.
Mistral Moderation
A classifier service for text content moderation.
Other models.
Leanstral
First open-source code agent for Lean 4.
Open
Mistral NeMo
State-of-the-art Mistral model trained specifically for code tasks.
Mistral Embed
State-of-the-art semantic model for extracting representation of text extracts.
