# Mistral AI

> European lab shipping open-weight models on its own API, every major cloud, or your GPUs.

Canonical: https://www.subconscious.dev/providers/mistral-ai · By The Subconscious Team · Updated September 30, 2026

- Founded: 2023
- Example models: Mistral Medium 3.5, Mistral Small 4
- Website: https://mistral.ai

## Overview

Mistral AI is a Paris lab that sells its models through La Plateforme, its own API, and releases most of them as open weights. It consolidated the lineup in 2026. Mistral Medium 3.5, released April 28, is a dense 128B model that merges instruction following, reasoning and coding into one set of weights, and it replaced both Devstral 2 and the Magistral reasoning models. It costs $1.50 in and $7.50 out per million tokens and scores 77.6% on SWE-Bench Verified by Mistral's count. Mistral Small 4, a 119B mixture-of-experts model with 6.5B active, costs $0.15 in and $0.60 out. Mistral Large 3, a 675B MoE under Apache 2.0, runs $0.50 in and $1.50 out. All three carry a 256K context window.

Around the general models, Mistral sells Codestral for low-latency fill-in-the-middle completion at $0.30 in and $0.90 out with 128K context, plus OCR, Voxtral speech models and an Agents API with built-in tools. Batch halves prices and cached input cuts input cost by up to 90%. The same models run on Azure AI, Amazon Bedrock, Vertex AI, Snowflake Cortex and IBM watsonx, and Medium 3.5 self-hosts on as few as four GPUs, with NVIDIA NIM containers available. Regional endpoints in Europe and the US went GA in August 2026 alongside a Priority Tier with uptime SLAs. The self-serve fine-tuning API is deprecated; custom training now goes through Forge, an enterprise system covering pre-training, post-training and RL.

## Upsides

- Open weights on the flagship models, so the same model can move from API to cloud to self-hosted.
- Low list prices, with Large 3 at $0.50 in and Small 4 at $0.15 in, plus 50% off on Batch.
- Choice of EU or US processing region, which suits teams with data residency rules.
- Available on Azure, Bedrock, Vertex AI, Snowflake and watsonx for cloud-credit buyers.

## Core use cases

- Agentic coding and long-horizon tool use on Medium 3.5.
- Self-hosted or sovereign deployments that need open weights and in-region processing.
- High-volume, cost-sensitive workloads on Small 4.

## Downsides

- Context tops out at 256K, well below the 1M windows offered by several US labs.
- Models retire fast: Devstral 2 and Magistral were deprecated within months of launch, which forces regular migrations.

## At a glance

| Attribute | Value |
|---|---|
| Model access | Open weights, plus closed Codestral |
| Flagship models | Mistral Medium 3.5, Small 4, Large 3 |
| Speed | - |
| Price | $0.15–$1.50 in, $0.60–$7.50 out per 1M |
| Customization | Forge (enterprise); fine-tuning API deprecated |
| Deployment | API, Azure, Bedrock, Vertex, self-host |
| Long context | 256K |

## FAQ

### What is Mistral AI?

Mistral AI is a Paris lab that sells its models through La Plateforme, its own API, and releases most of them as open weights. It consolidated the lineup in 2026. Mistral Medium 3.5, released April 28, is a dense 128B model that merges instruction following, reasoning and coding into one set of weights, and it replaced both Devstral 2 and the Magistral reasoning models. It costs $1.50 in and $7.50 out per million tokens and scores 77.6% on SWE-Bench Verified by Mistral's count. Mistral Small 4, a 119B mixture-of-experts model with 6.5B active, costs $0.15 in and $0.60 out. Mistral Large 3, a 675B MoE under Apache 2.0, runs $0.50 in and $1.50 out. All three carry a 256K context window.

### What is Mistral AI best for?

Agentic coding and long-horizon tool use on Medium 3.5; Self-hosted or sovereign deployments that need open weights and in-region processing; High-volume, cost-sensitive workloads on Small 4.

### How much does Mistral AI cost?

Mistral AI pricing at a glance: $0.15–$1.50 in, $0.60–$7.50 out per 1M. Rates change often, so check Mistral AI's pricing page before committing.

### How much context does Mistral AI support?

Mistral AI's long-context support: 256K.

### What are the downsides of Mistral AI?

Context tops out at 256K, well below the 1M windows offered by several US labs; Models retire fast: Devstral 2 and Magistral were deprecated within months of launch, which forces regular migrations.

### What are the best alternatives to Mistral AI?

Common alternatives include Subconscious, OpenAI, Anthropic, Google Vertex AI, Amazon Bedrock. Each has a head-to-head comparison with Mistral AI on this site.

## Comparisons

- [Subconscious vs Mistral AI](https://www.subconscious.dev/compare/subconscious-vs-mistral-ai.md)
- [OpenAI vs Mistral AI](https://www.subconscious.dev/compare/openai-vs-mistral-ai.md)
- [Anthropic vs Mistral AI](https://www.subconscious.dev/compare/anthropic-vs-mistral-ai.md)
- [Google Vertex AI vs Mistral AI](https://www.subconscious.dev/compare/google-vertex-vs-mistral-ai.md)
- [Amazon Bedrock vs Mistral AI](https://www.subconscious.dev/compare/aws-bedrock-vs-mistral-ai.md)
- [Together AI vs Mistral AI](https://www.subconscious.dev/compare/together-ai-vs-mistral-ai.md)
- [Fireworks AI vs Mistral AI](https://www.subconscious.dev/compare/fireworks-vs-mistral-ai.md)
- [Baseten vs Mistral AI](https://www.subconscious.dev/compare/baseten-vs-mistral-ai.md)
- [Groq vs Mistral AI](https://www.subconscious.dev/compare/groq-vs-mistral-ai.md)
- [Cerebras vs Mistral AI](https://www.subconscious.dev/compare/cerebras-vs-mistral-ai.md)
- [DeepInfra vs Mistral AI](https://www.subconscious.dev/compare/deepinfra-vs-mistral-ai.md)
- [Hugging Face Inference Providers vs Mistral AI](https://www.subconscious.dev/compare/hugging-face-vs-mistral-ai.md)
- [Modal vs Mistral AI](https://www.subconscious.dev/compare/modal-vs-mistral-ai.md)
- [Cloudflare Workers AI vs Mistral AI](https://www.subconscious.dev/compare/cloudflare-workers-ai-vs-mistral-ai.md)
- [xAI vs Mistral AI](https://www.subconscious.dev/compare/xai-vs-mistral-ai.md)
- [Mistral AI vs DeepSeek](https://www.subconscious.dev/compare/mistral-ai-vs-deepseek.md)
- [Mistral AI vs Moonshot AI](https://www.subconscious.dev/compare/mistral-ai-vs-moonshot-ai.md)
- [Mistral AI vs Z.ai](https://www.subconscious.dev/compare/mistral-ai-vs-z-ai.md)
- [Mistral AI vs Alibaba Cloud](https://www.subconscious.dev/compare/mistral-ai-vs-alibaba-cloud.md)
- [Mistral AI vs Meta](https://www.subconscious.dev/compare/mistral-ai-vs-meta.md)
- [Mistral AI vs Cohere](https://www.subconscious.dev/compare/mistral-ai-vs-cohere.md)
- [Mistral AI vs SambaNova](https://www.subconscious.dev/compare/mistral-ai-vs-sambanova.md)
- [Mistral AI vs Nebius](https://www.subconscious.dev/compare/mistral-ai-vs-nebius.md)
- [Mistral AI vs Crusoe](https://www.subconscious.dev/compare/mistral-ai-vs-crusoe.md)
- [Mistral AI vs fal](https://www.subconscious.dev/compare/mistral-ai-vs-fal.md)
- [Mistral AI vs Novita AI](https://www.subconscious.dev/compare/mistral-ai-vs-novita-ai.md)
- [Mistral AI vs Venice](https://www.subconscious.dev/compare/mistral-ai-vs-venice.md)
- [Mistral AI vs Parasail](https://www.subconscious.dev/compare/mistral-ai-vs-parasail.md)
- [Mistral AI vs Inference.net](https://www.subconscious.dev/compare/mistral-ai-vs-inference-net.md)
- [Mistral AI vs GMI Cloud](https://www.subconscious.dev/compare/mistral-ai-vs-gmi-cloud.md)
- [Mistral AI vs Thinking Machines](https://www.subconscious.dev/compare/mistral-ai-vs-thinking-machines.md)
- [Mistral AI vs Sail Research](https://www.subconscious.dev/compare/mistral-ai-vs-sail-research.md)
- [Mistral AI vs Morph](https://www.subconscious.dev/compare/mistral-ai-vs-morph.md)
- [Mistral AI vs Relace](https://www.subconscious.dev/compare/mistral-ai-vs-relace.md)
- [Mistral AI vs TypeSafe AI](https://www.subconscious.dev/compare/mistral-ai-vs-typesafe-ai.md)
- [Mistral AI vs StepFun](https://www.subconscious.dev/compare/mistral-ai-vs-stepfun.md)
- [Mistral AI vs Runware](https://www.subconscious.dev/compare/mistral-ai-vs-runware.md)
- [Mistral AI vs StreamLake](https://www.subconscious.dev/compare/mistral-ai-vs-streamlake.md)
- [Mistral AI vs Wafer](https://www.subconscious.dev/compare/mistral-ai-vs-wafer.md)
- [Mistral AI vs RunInfra](https://www.subconscious.dev/compare/mistral-ai-vs-runinfra.md)
- [Mistral AI vs Particle.AI](https://www.subconscious.dev/compare/mistral-ai-vs-particle-ai.md)

## Sources

- [Mistral pricing](https://mistral.ai/pricing)
- [Mistral Medium 3.5 launch](https://mistral.ai/news/vibe-remote-agents-mistral-medium-3-5/)
- [Mistral models overview](https://docs.mistral.ai/getting-started/models/models_overview/)
- [Mistral regional inference](https://mistral.ai/news/regional-inference-open-models-new-compute/)

Pricing and model lineups change often; figures are a snapshot.