• Models

It starts with frontier science.

State-of-the-art models with cutting edge capabilities, from cloud to edge.

Get in touch

Codestral

Devstral 2

Mistral Medium 3.5

Mistral Small 4

OCR 4.1

Voxtral TTS

Ministral 3 (14B)

Mistral Moderation 2

Codestral Embed

Leanstral 1.5

Classifier API model (3B)

OCR 4.1

The world’s best document extraction and understanding model.

Learn more

OCR

Multimodal

Text-to-text

Mistral Medium 3.5

Merged dense model unifying instruction, reasoning, and coding; configurable reasoning effort with multimodal input.

Learn more

Text-to-text

Agentic

Multimodal

Coding

Reasoning

Voxtral TTS

Speech generation with reference voices.

Learn more

Voice

Model families.

Generalist models.

Apache 2.0

Mistral Large 3

Multimodal mixture-of-experts architecture; base and instruct releases.

Text-to-text

Multimodal

Modified MIT

Mistral Medium 3.5

Merged dense model unifying instruction, reasoning, and coding; configurable reasoning effort with multimodal input.

Text-to-text

Agentic

Multimodal

Coding

Reasoning

Learn more

Apache 2.0

Mistral Small 4

Hybrid instruct and reasoning MoE; unifies Instruct, Magistral, and Devstral into one model with multimodal input.

Text-to-text

Agentic

Multimodal

Lightweight

Read more

Open

Ministral 3 (8B)

Best-in-class frontier AI to the edge.

Text-to-text

Agentic

Lightweight

Specialist models.

Premier

OCR 4.1

The world’s best document extraction and understanding model.

OCR

Multimodal

Text-to-text

Read more

CC BY-NC 4.0

Voxtral TTS

Speech generation with reference voices.

Voice

Read more

Premier

Voxtral Mini Transcribe 2

State-of-the-art batch audio transcription.

Transcription

Voice

Lightweight

Apache 2.0

Voxtral Mini Transcribe Realtime

Streaming transcription with configurable latency.

Transcription

Voice

Learn more

Apache 2.0

Voxtral Small

Audio and text understanding; transcribes, translates, and answers questions from speech with text output.

Transcription

Voice

Text-to-text

Open

Devstral 2

Open-weights agentic coding model for autonomous software engineering.

Coding

Agentic

Text-to-text

Read more

Premier

Codestral

Low-latency coding model optimized for high-frequency completion, fill-in-the-middle, and code generation tasks.

Coding

Text-to-text

Premier

Codestral Embed

State-of-the-art embeddings for code and natural language queries.

Embedding

Coding

Mistral Moderation 2

A classifier service for text content moderation.

Classifier APIs

Other models.

Apache 2.0

Leanstral 1.5

Proof engineering and verification in Lean 4.

Coding

Proof engineering

44e33e3e-0fd8-43d7-957e-d2689d4c9313

Open

Mistral NeMo

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

Coding

Lightweight

Mistral Embed

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

Embedding

Text-to-text

Need domain-specialized models?

Collaborate with Mistral to customize models with your data, aligning performance with your enterprise’s unique requirements.

Learn more about model customization

Ready to get started?

Explore the full lineup of models, compare benchmarks, and find the right model for your use case.

sitting

Go to docs

gradient background
cat sitting