---
title: "EU LLM API · Claude, Gemini and open models in Europe"
description: "Claude, Gemini, open-weight LLMs, embeddings and speech models through one API, processed in EU data centres. Switch from OpenAI by changing one line."
canonical: "https://inference.eu/models/"
language: "en"
---

LLM API hosted in the EU

# An EU LLM API for developers, serving Claude, Gemini and open models

Send your prompts to Claude, Gemini and open-weight models running in European data centres instead of a US provider. The API uses the OpenAI format, and accepts the Anthropic Messages format too, so your existing code works after you change the base URL and key.

[Contact sales](https://inference.eu/contact/)[Every model, one EU endpoint](#directory)

## How it works

-   ### Works with the OpenAI SDK

    Python, JavaScript, LangChain, LlamaIndex, the Vercel AI SDK and anything else that speaks the OpenAI format. Change the base URL and key; nothing else.

-   ### Speaks the Anthropic format too

    Clients written for the Anthropic Messages API call the same open-weight models. Point them at the EU endpoint and pick a model.

-   ### Switch models with one string

    Every model sits behind the same endpoint. Try a different one by changing the model name in the request.

-   ### Requests processed in the EU

    Prompts and responses are processed in European data centres, and failover moves traffic to another European location.

-   ### Never used for training

    What you send is used to answer the request and nothing else. It never goes into a training run.

-   ### Chat, code, embeddings and speech

    Text and code generation, embeddings and reranking for search, speech to text and image models, all with one key.

-   ### Priced per token, billed in euros

    Pay for the tokens you use, on one invoice covering every model.

## Every model, one EU endpoint

Claude, Gemini and open-weight models for chat, code, embeddings, reranking, speech and images. Switch between them by changing the model string.

### Chat

-   [Apertus 1.5 70BRun Apertus 1.5 70B, the fully open Swiss model from EPFL and ETH Zurich, on inference.eu with an OpenAI-compatible API and EU data processing.Swiss AI Initiative (EPFL, ETH Zurich, CSCS) · 70B](https://inference.eu/models/apertus-70b/)
-   [Claude Haiku 5.5Use Anthropic's Claude Haiku 5.5 on inference.eu for high-volume tasks: requests processed in EU data centres, one DPA, nothing used for training.Anthropic](https://inference.eu/models/claude-haiku-5-5/)
-   [Claude Sonnet 5.5Use Anthropic's Claude Sonnet 5.5 on inference.eu: requests processed in EU data centres, one data processing agreement, nothing used for training.Anthropic](https://inference.eu/models/claude-sonnet-5-5/)
-   [Clef FlashRun Clef Flash on inference.eu: a 9B decision model that returns probabilities for typed questions instead of free text, processed in the EU.Cloudflare · 9B](https://inference.eu/models/clef-flash/)
-   [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.UTTER project (Instituto Superior Técnico, Unbabel, University of Edinburgh and partners) · 22.6B](https://inference.eu/models/eurollm-22b-instruct/)
-   [Gemini 3.8 FlashUse Google's Gemini 3.8 Flash on inference.eu: a fast model with 1M context, requests processed in EU data centres and nothing used for training.Google DeepMind](https://inference.eu/models/gemini-3-8-flash/)
-   [Llama 3.3 70B InstructRun Meta's Llama 3.3 70B Instruct on inference.eu: an OpenAI-compatible API with requests processed in EU data centres and billed in euros.Meta · 70B](https://inference.eu/models/llama-3-3-70b-instruct/)
-   [Mistral Small 4 119BRun Mistral Small 4 119B from French lab Mistral AI on inference.eu: reasoning, vision and code in one model, processed in EU data centres.Mistral AI · 119B](https://inference.eu/models/mistral-small-4-119b/)
-   [Qwen3 235B A22B InstructRun Qwen3 235B A22B Instruct (2507) on inference.eu: a 256K-context open model behind an OpenAI-compatible API, processed in EU data centres.Qwen (Alibaba Cloud) · 235B](https://inference.eu/models/qwen3-235b-a22b-instruct/)
-   [Teuken 7B InstructRun Teuken 7B Instruct, the German open model trained on all 24 official EU languages, on inference.eu through an OpenAI-compatible API.OpenGPT-X (Fraunhofer, Forschungszentrum Jülich, TU Dresden, DFKI) · 7B](https://inference.eu/models/teuken-7b-instruct/)

### Reranking

-   [BGE Reranker v2 M3Run bge-reranker-v2-m3 on inference.eu: a light multilingual reranker that sorts search and RAG results by relevance, processed in the EU.BAAI (Beijing Academy of Artificial Intelligence) · 568M](https://inference.eu/models/bge-reranker-v2-m3/)

### Embeddings

-   [BGE-M3Run bge-m3 on inference.eu: a compact multilingual embedding model for search and RAG, 100+ languages and 8192-token inputs, processed in the EU.BAAI (Beijing Academy of Artificial Intelligence) · 568M](https://inference.eu/models/bge-m3/)
-   [multilingual-e5-large-instructRun multilingual-e5-large-instruct on inference.eu: compact embeddings in 100 languages for search, RAG and classification, processed in the EU.Microsoft (intfloat) · 560M](https://inference.eu/models/multilingual-e5-large-instruct/)
-   [Qwen3 Embedding 8BRun Qwen3 Embedding 8B on inference.eu: multilingual text embeddings for search and RAG, up to 4096 dimensions and 32K tokens, processed in the EU.Qwen (Alibaba Cloud) · 8B](https://inference.eu/models/qwen3-embedding-8b/)

### Reasoning

-   [Claude Opus 5Use Anthropic's Claude Opus 5 on inference.eu: requests processed in EU data centres, one data processing agreement, and nothing used for training.Anthropic](https://inference.eu/models/claude-opus-5/)
-   [DeepSeek V4 FlashRun DeepSeek V4 Flash (0731) on inference.eu: a 1M-token open reasoning model behind an OpenAI-compatible API, processed in EU data centres, not in China.DeepSeek · 284B](https://inference.eu/models/deepseek-v4-flash/)
-   [DeepSeek V4.1 Flash (1017)Run DeepSeek V4.1 Flash (1017) on inference.eu: a multimodal open reasoning model with 1M context, OpenAI-compatible, processed in the EU, not in China.DeepSeek · 552B](https://inference.eu/models/deepseek-v4-1-flash-1017/)
-   [Gemini 4 ArgonUse Google's Gemini 4 Argon on inference.eu: requests processed in EU data centres, one data processing agreement, and nothing used for training.Google DeepMind](https://inference.eu/models/gemini-4-argon/)
-   [GLM-5.3Run Z.ai's GLM-5.3, an open-weight model built for complex coding and long agent tasks, on inference.eu with OpenAI and Anthropic APIs, processed in the EU.Z.ai (Zhipu AI) · 753B](https://inference.eu/models/glm-5-3/)
-   [GLM-5.3-FlashRun Z.ai's GLM-5.3-Flash on inference.eu: a multimodal open model with 18B active parameters and 1M context, OpenAI-compatible, processed in the EU.Z.ai (Zhipu AI) · 320B](https://inference.eu/models/glm-5-3-flash/)
-   [gpt-oss-120bRun OpenAI's open-weight gpt-oss-120b reasoning model on inference.eu: the OpenAI API format, with requests processed in EU data centres, billed in euros.OpenAI · 117B](https://inference.eu/models/gpt-oss-120b/)
-   [gpt-oss-20bRun OpenAI's compact open-weight gpt-oss-20b reasoning model on inference.eu: the OpenAI API format, processed in EU data centres, billed in euros.OpenAI · 21B](https://inference.eu/models/gpt-oss-20b/)
-   [Kimi K2.6Run Moonshot AI's Kimi K2.6 on inference.eu: a 1-trillion-parameter open agentic model with image input and 256K context, processed in EU data centres.Moonshot AI · 1T](https://inference.eu/models/kimi-k2-6/)
-   [Kimi K3Run Kimi K3, Moonshot AI's 2.8-trillion-parameter open-weight model with 1M context and image and video input, on inference.eu with processing in the EU.Moonshot AI · 2.8T](https://inference.eu/models/kimi-k3/)
-   [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.Mistral AI · 24B](https://inference.eu/models/magistral-small/)
-   [MiniMax M3Run MiniMax M3 on inference.eu: an open multimodal model with 1M context for coding and agent work, OpenAI-compatible and processed in EU data centres.MiniMax · 428B](https://inference.eu/models/minimax-m3/)
-   [Qwen3.8 27BRun Qwen3.8 27B on inference.eu: a dense open model with image and video input, adjustable reasoning and 256K context, OpenAI-compatible, processed in the EU.Qwen (Alibaba Cloud) · 27B](https://inference.eu/models/qwen3-8-27b/)

### OCR and documents

-   [DeepSeek-OCR 2Run DeepSeek-OCR 2 on inference.eu: a 3B open OCR model that turns scans and PDFs into Markdown, processed in EU data centres, not in China.DeepSeek · 3B](https://inference.eu/models/deepseek-ocr-2/)

### Code

-   [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.Mistral AI · 24B](https://inference.eu/models/devstral-small-2/)
-   [Qwen3.6 27BRun Qwen3.6 27B on inference.eu: a dense open model for agentic coding with image input and 256K context, OpenAI-compatible and processed in the EU.Qwen (Alibaba Cloud) · 27B](https://inference.eu/models/qwen3-6-27b/)
-   [Qwen3.6 35B A3BRun Qwen3.6 35B A3B on inference.eu: a fast mixture-of-experts model with 3B active parameters, image input and 256K context, processed in the EU.Qwen (Alibaba Cloud) · 35B](https://inference.eu/models/qwen3-6-35b-a3b/)

### Vision

-   [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.Google DeepMind · 25.2B](https://inference.eu/models/gemma-4-26b-a4b/)
-   [Gemma 4 31BRun Google's Gemma 4 31B on inference.eu: an open multimodal model with 256K context behind an OpenAI-compatible API, processed in EU data centres.Google DeepMind · 30.7B](https://inference.eu/models/gemma-4-31b/)

### Speech to text

-   [Voxtral Small 24BTranscribe and understand audio with Voxtral Small 24B from Mistral AI on inference.eu: an OpenAI-compatible API with processing in EU data centres.Mistral AI · 24B](https://inference.eu/models/voxtral-small-24b/)
-   [Whisper Large V3 TurboTranscribe audio with OpenAI's open Whisper Large V3 Turbo on inference.eu: the OpenAI transcription API format, with recordings processed in the EU.OpenAI · 809M](https://inference.eu/models/whisper-large-v3-turbo/)

## Switch from OpenAI by changing one line

The API uses the OpenAI request format, so the SDKs and frameworks you already use work as they are. Change the base URL and the key, then compare models on your own prompts.

```
from openai import OpenAI

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

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

```
import OpenAI from "openai"

const client = new OpenAI({
  baseURL: "https://api.inference.eu/v1",
  apiKey: process.env.INFERENCE_EU_API_KEY,
})

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

```
curl https://api.inference.eu/v1/chat/completions \
  -H "Authorization: Bearer $INFERENCE_EU_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "mistral-small-4-119b",
    "messages": [{"role": "user", "content": "Summarise this contract in three bullet points"}]
  }'
```

## What teams use it for

-   ### AI features in a SaaS product

    Add chat, summaries or search to your product without sending your customers' data outside the EU.

-   ### Search over internal documents

    Embed contracts, tickets and wikis, then answer questions about them with retrieval-augmented generation.

-   ### Moving off a US AI provider

    Keep your prompts and code, point them at an EU endpoint, and compare quality model by model.

-   ### Agents and coding tools

    Run coding assistants and agents on open-weight models in the EU, through the OpenAI or Anthropic format.

## Models: common questions

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

Open-weight models for chat, reasoning, code, embeddings, reranking, speech and images, from European and international labs. Each model page lists its maker, licence and context length, and the request id to use. New open-weight releases are added to the same endpoint, so you can try them without changing anything else.

No. If you use an OpenAI SDK, or a framework built on one, change the base URL and the API key. Prompts, tools and streaming work as before, and you choose a model by its name in the request, just as you do today.

Yes. The API also accepts requests in the Anthropic Messages format, so clients and SDKs written for Claude work unchanged, with Claude itself or any open-weight model hosted in the EU. Set the base URL and key in the client, then choose a model by its name.

No. Requests are processed to produce your response and are not used to train any model. The same commitment is written into the data processing agreement you sign with us, so it is a contractual term rather than a promise on a web page.

In data centres located in the European Union. If a location fails, traffic moves to another European location rather than to a region outside the EU. You get the list of locations and subprocessors with your data processing agreement.

That depends on your data and your purpose, which only you can assess. We give your data protection officer what they need to do it, with processing in the EU, a data processing agreement under GDPR Article 28, a full list of subprocessors and a contractual commitment not to train on your data.

Your requests stay in European data centres instead of going to the United States. The models are open-weight, with published weights, so you are never tied to one company's model. Billing is in euros, and one data processing agreement covers every model on the platform.

You pay for what you use, measured in tokens for language models, and receive one invoice in euros covering every model and every API key in your workspace. You don't need a separate contract or account with each model maker.

Yes. Embedding models turn your documents into vectors for a search index, and rerank models sort the retrieved passages by relevance before they reach a chat model. Both run behind the same API and key, so a whole retrieval pipeline stays in the EU.

Yes. Speech-to-text models transcribe audio, text-to-speech models read text aloud, and image models generate pictures from a text prompt. They use the same key and the same invoice as the language models, so you don't need a second provider for audio or images.

An open-weight model publishes its trained weights so anyone can run it, often under a licence with conditions. Open-source in the strict sense also covers the training code and data. Each model page links to its licence, so you can check what commercial use it allows.

Start with one workload. Point your existing code at the EU base URL, pick an open-weight model of similar size and compare outputs on your own test prompts. Expect to adjust a prompt or two, then move the remaining traffic once the quality matches what you need.

Yes, on a dedicated GPU server that you rent by the hour in a European data centre. You serve your own weights there, behind an endpoint you control, and the hours appear on the same invoice in euros as your API usage.

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

## 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