---
title: "Mistral Small 4 API hosted in the EU"
description: "Run Mistral Small 4 119B from French lab Mistral AI on inference.eu: reasoning, vision and code in one model, processed in EU data centres."
canonical: "https://inference.eu/models/mistral-small-4-119b/"
language: "en"
---

Chat · Mistral AI

# Run Mistral Small 4 in the EU through an OpenAI-compatible API

Mistral Small 4 is a 119-billion-parameter mixture-of-experts model from Paris-based Mistral AI that combines instruction following, reasoning and coding in one set of weights. On inference.eu you call it with the OpenAI or Anthropic request format, with requests processed in EU data centres and never used for training.

[Contact sales](https://inference.eu/contact/)[Example request](#example)

1.  01119B · 6.5B
2.  02EU data centres
3.  03Apache 2.0

## What Mistral Small 4 is

Mistral Small 4, published as Mistral-Small-4-119B-2603, is a mixture-of-experts model with 128 experts, four of which are active per token. It has 119 billion parameters in total and about 6.5 billion active per token, with a context window of 256K tokens and image input alongside text.

Mistral describes it as a merger of three model families that used to be separate: the general Instruct models, the Magistral reasoning models and the Devstral coding models. You choose how much it reasons per request, so one model can handle quick replies and harder multi-step problems. The weights are released under the Apache 2.0 licence.

## What it does well, and where it falls short

Mistral Small 4 is a strong choice for European companies that want one versatile model: it handles chat, document analysis, coding help, image understanding and agent tool calls. The model card lists support for more than twenty languages, including English, French, German, Spanish, Italian, Portuguese, Polish, Romanian and Swedish, and it follows system prompts closely.

With only about 6.5 billion active parameters per token it is efficient to serve, but it does not reach the depth of the very large open models such as GLM-5.3 or Kimi K3 on long, hard agentic coding tasks. Reasoning mode adds output tokens and time, so switch it off for simple requests. It reads images but does not produce them.

## Calling Mistral Small 4 through the API

inference.eu uses the OpenAI request format, so the OpenAI SDKs, LangChain, LlamaIndex, the Vercel AI SDK and the Mistral-compatible tools that accept a custom base URL all work. Set the base URL to the inference.eu API, use your inference.eu key and set the model to `mistral-small-4-119b`.

Send images as `image_url` content parts, just as you would with OpenAI vision models. Use the `reasoning_effort` parameter to ask for more or less reasoning on each request. Tool calls, streaming and JSON output work as usual, and the Anthropic Messages format is also available for tools built around Claude.

## A French model, processed in the EU

Mistral Small 4 is one of the few strong open models built in Europe. On inference.eu it also runs in Europe: your prompts, images and responses are processed in data centres in the European Union, and they are never used to train any model.

We sign a data processing agreement under Article 28 of the GDPR with every customer and share our subprocessor list, so you can document where data goes for your records of processing. Billing is in euros. The Apache 2.0 licence allows commercial use without user limits or revenue thresholds, which keeps the legal review short.

## Mistral Small 4 119B

Maker

Mistral AI

Made in

France

Type

Chat

Parameters

119B

Active parameters

6.5B

Context window

262,144 tokens

Licence

[Apache 2.0](https://www.apache.org/licenses/LICENSE-2.0)

Released

March 2026

Model ID

mistral-small-4-119b

Processing

EU data centres

Training on your data

Never

Source

[huggingface.co](https://huggingface.co/mistralai/Mistral-Small-4-119B-2603)

mistral-small-4-119b · Python

```
import os
from openai import OpenAI

client = OpenAI(base_url="https://api.inference.eu/v1", api_key=os.environ["INFERENCE_EU_API_KEY"])

response = client.chat.completions.create(
    model="mistral-small-4-119b",
    messages=[{"role": "user", "content": "Summarise this contract in five bullet points."}],
)
print(response.choices[0].message.content)
```

## What teams use it for

-   1

    ### One model for a whole product

    Use the same model for chat, summaries, image questions and tool calls instead of maintaining separate models for each task.

-   2

    ### Document and screenshot analysis

    Read scanned forms, charts and screenshots together with text instructions and return structured answers.

-   3

    ### European-language assistants

    Build assistants for French, German, Spanish, Italian, Polish or Romanian users on a model trained by a European lab.

## Mistral Small 4 119B: common questions

The questions buyers, developers and legal teams ask us most. Something missing? Ask us directly.

GDPR compliance depends on how you use the model. On inference.eu, Mistral Small 4 requests are processed in EU data centres, not used for training, and covered by a data processing agreement under Article 28 GDPR with a published subprocessor list. That gives you the basis to process personal data with it.

Mistral Small 4 supports a context window of 256K tokens. That is enough for long reports, several contracts or a sizeable part of a codebase in one request. Images count towards the context too, so very large images reduce the room left for text.

Yes. Mistral Small 4 is released under the Apache 2.0 licence, which allows commercial use, modification and redistribution with no user or revenue limits. This differs from some larger Mistral models, which use more restrictive licences.

Mistral says Small 4 unifies its Instruct, Magistral reasoning and Devstral coding families in one model. For most new projects it is the simpler choice. Magistral Small and Devstral Small 2 remain useful when you want a smaller dense 24B model specialised for one job.

Yes. Mistral Small 4 accepts images alongside text, so it can describe photos, read charts and extract information from screenshots or scanned pages. It produces text only. Send images through the standard OpenAI \`image\_url\` content format.

Pass the \`reasoning\_effort\` parameter in your request. A higher setting makes the model think through the problem before answering, which helps with maths, planning and code. For simple chat or extraction, a low setting keeps answers short and quick.

Both are efficient mixture-of-experts models with image input and 256K context. Mistral Small 4 is larger in total, built in France and combines reasoning and coding modes. Gemma 4 26B A4B is smaller, with about 3.8 billion active parameters. Test both on your own data.

The model card lists more than twenty languages, including English, French, German, Spanish, Portuguese, Italian, Polish, Romanian, Swedish, Ukrainian, Russian, Arabic, Chinese, Japanese and Korean. For smaller EU languages such as Estonian or Maltese, EuroLLM covers all official EU languages.

Change the base URL of your OpenAI client to the inference.eu API, use your inference.eu key and set the model to \`mistral-small-4-119b\`. Messages, images, tools and streaming keep working with the same code.

No. Mistral AI publishes the weights; it is not involved when you call the model on inference.eu. Your requests are processed in EU data centres to produce your response and are never used to train Mistral Small 4 or any other model.

[Ask a question](https://inference.eu/contact/)

## Related

[All models](https://inference.eu/models/#directory)

-   [Magistral Small 1.2Run Magistral Small 1.2, Mistral AI's open 24B reasoning model, on inference.eu through an OpenAI-compatible API with processing in EU data centres.Reasoning](https://inference.eu/models/magistral-small/)
-   [Devstral Small 2Run Devstral Small 2, Mistral AI's open 24B coding agent model, on inference.eu with OpenAI and Anthropic APIs and processing in EU data centres.Code](https://inference.eu/models/devstral-small-2/)
-   [Gemma 4 26B A4BRun Google's Gemma 4 26B A4B on inference.eu: a fast mixture-of-experts model with image input and 256K context, processed in EU data centres.Vision](https://inference.eu/models/gemma-4-26b-a4b/)
-   [EuroLLM 22B InstructRun EuroLLM 22B Instruct, the European open model trained for all 24 official EU languages, on inference.eu with an OpenAI-compatible API.Chat](https://inference.eu/models/eurollm-22b-instruct/)

## Move your AI workloads to EU infrastructure

Tell us which models you use and roughly how much traffic you have. We reply within one working day with pricing and next steps.

-   Reply within one working day
-   Pricing for your volume
-   DPA ready to sign