frontieropen-sourcereasoning

Mistral AI

European AI lab pioneering efficient open and commercial frontier models.

4
Models tracked
2023
Founded
Mistral Large 2
Flagship model
Paris
HQ
Visit mistral.aiDocumentation

About

About Mistral AI

Mistral AI, founded in Paris in 2023, consistently delivers outsized performance at smaller parameter counts. Their open models (Mistral 7B, Mixtral 8x7B) disrupted the market with efficiency; their commercial line (Mistral Large) competes with GPT-4 class models. Codestral and Devstral are specialised coding models. Mistral operates La Plateforme API and offers open-weight downloads for self-hosting.

mistral.ai

Key highlights

  • Open-weight models available for commercial self-hosting

  • Codestral with 256K context purpose-built for code completion

  • Mistral 7B and Mixtral sparked the efficient open-model movement in 2023

  • Devstral leads open models on SWE-bench Verified for agentic coding

  • La Plateforme API with European data residency options

Models

4 models tracked

Context windows, pricing, capabilities, and API identifiers — everything you need to choose the right model for your workload.

Flagship

Mistral Large 2

2024

128Kcontext
TextCodeReasoning

Mistral's flagship commercial model. A 123B parameter model with strong performance across reasoning, code, and multilingual tasks. Supports 80+ languages natively.

  • 80+ language support
  • Strong tool use
  • Function calling
  • Reliable reasoning
In: $2 / M tokens
Out: $6 / M tokens
mistral-large-latest
Open Source

Mistral Small 3.1

2025

128Kcontext
TextVisionCode

A multimodal open-weight model with vision capability and fast inference at a fraction of Large's cost. Available for self-hosting under the Mistral Research Licence.

  • Vision capable
  • Open weights (MRL)
  • Very low cost
  • Fast inference
In: $0.10 / M tokens
Out: $0.30 / M tokens
mistral-small-latest

Codestral

2024

256Kcontext
Code

Purpose-built for code completion and generation. Supports fill-in-the-middle (FIM) mode for VS Code and JetBrains plugins, with a 256K context for large codebases.

  • Fill-in-the-middle (FIM)
  • 256K code context
  • 80+ languages
  • IDE plugins (VS Code, JetBrains)
In: $0.20 / M tokens
Out: $0.60 / M tokens
codestral-latest
Open Source

Devstral

2025

128Kcontext
CodeReasoning

An open-source agentic coding model jointly developed with All Hands AI. Leads open-source models on SWE-bench Verified for autonomous software engineering.

  • Agentic software engineering
  • SWE-bench leader (open source)
  • Tool use
  • Open weights
devstral-latest

Use Cases

What Mistral AI is best for

Each provider has specialisms shaped by their architecture decisions, training data, and API design. Here is where Mistral AI performs best.

Multilingual customer support
Code completion in IDEs
On-premise enterprise deployments
European data-residency requirements
Agentic software engineering

Get started

Start building with Mistral AI

Visit the official Mistral AI platform to access API keys, documentation, and pricing details.

Visit mistral.aiRead the docsBack to all providers