---
title: "Llama 3.3 70B API hosted in the EU"
description: "Run 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."
canonical: "https://inference.eu/models/llama-3-3-70b-instruct/"
language: "en"
---

Chat · Meta

# Run Llama 3.3 70B in the EU through an OpenAI-compatible API

Llama 3.3 70B Instruct is Meta's 70-billion-parameter text model, released in December 2024 and still one of the most widely deployed open models. On inference.eu you call it with the OpenAI or Anthropic request format, your requests are processed in EU data centres, and nothing you send is used for training.

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

1.  0170B
2.  02EU data centres
3.  03Llama 3.3 Community License

## What Llama 3.3 70B is

Llama 3.3 70B Instruct is a dense, text-only language model from Meta with 70 billion parameters and a context window of 128K tokens. It is an instruction-tuned model, built on Llama 3.1 70B, and Meta lists English, German, French, Italian, Portuguese, Spanish, Hindi and Thai as supported languages.

Because it is dense rather than a mixture of experts, every token passes through all 70 billion parameters. That makes its behaviour predictable and well understood: it has been tested by thousands of teams since its release, and most frameworks, prompt libraries and evaluation tools support it out of the box.

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

Llama 3.3 70B is a solid all-rounder for chat, summarisation, classification, extraction and drafting in the major Western European languages. It follows system prompts reliably, supports tool calling and produces consistent JSON when you ask for it, which is why many teams use it as the default model behind internal assistants.

It is no longer the strongest open model. Newer releases such as Qwen3, Gemma 4 or Mistral Small 4 reason better, write better code and accept images, which Llama 3.3 does not. It has no separate thinking mode, so multi-step maths and complex agent tasks are better served by a reasoning model. Choose it when you want a proven, stable baseline.

## Calling Llama 3.3 70B through the API

inference.eu speaks the OpenAI request format, so the official OpenAI SDKs for Python and JavaScript, LangChain, LlamaIndex and the Vercel AI SDK work unchanged. Set the base URL to the inference.eu API, use your inference.eu key, and set the model to `llama-3.3-70b-instruct`.

Chat completions, streaming, tool calls and JSON mode work as you would expect. If your code uses the Anthropic Messages format instead, the same model is available there too, so tools built for Claude can switch to Llama 3.3 without a rewrite. Your prompts, few-shot examples and output parsing stay exactly as they are.

## Data residency for Llama 3.3 70B

When you call Llama 3.3 70B on inference.eu, your prompts and the model's responses are processed in data centres in the European Union. They are used to answer your request and are never used to train any model.

We sign a data processing agreement under Article 28 of the GDPR with every customer and provide our list of subprocessors, so your data protection officer can see exactly who handles the data and where. Billing is in euros. Note that the Llama 3.3 Community License has its own terms, including an acceptable use policy and an attribution requirement, which apply to how you use the model.

## Llama 3.3 70B Instruct

Maker

Meta

Made in

United States

Type

Chat

Parameters

70B

Context window

131,072 tokens

Licence

[Llama 3.3 Community License](https://www.llama.com/llama3_3/license/)

Released

December 2024

Model ID

llama-3.3-70b-instruct

Processing

EU data centres

Training on your data

Never

Source

[huggingface.co](https://huggingface.co/meta-llama/Llama-3.3-70B-Instruct)

llama-3.3-70b-instruct · 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="llama-3.3-70b-instruct",
    messages=[{"role": "user", "content": "Summarise this contract in five bullet points."}],
)
print(response.choices[0].message.content)
```

## What teams use it for

-   1

    ### Internal assistants

    A dependable default for company chat assistants that answer questions, rewrite text and summarise documents in English, German, French, Italian and Spanish.

-   2

    ### Structured extraction

    Pull names, dates, amounts and clauses out of emails, invoices and contracts into JSON that your systems can process.

-   3

    ### Baseline for migrations

    A well-known reference model when you move a workload from a US provider and want to compare quality model by model.

## Llama 3.3 70B Instruct: common questions

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

Compliance depends on how you use a model, not on the model alone. On inference.eu, requests are processed in EU data centres, are not used for training, and are covered by a data processing agreement under GDPR Article 28 with a published subprocessor list. That gives you the contractual basis to use Llama 3.3 with personal data.

Llama 3.3 70B Instruct supports a context window of 128K tokens, which is 131,072 tokens. That is enough for long contracts, full support threads or several reports in a single request. Prompt and response together must fit inside the window.

Yes. The Llama 3.3 Community License allows commercial use. Companies whose products had more than 700 million monthly active users when Llama 3.3 was released need a separate licence from Meta. You must also follow Meta's acceptable use policy and show the attribution the licence asks for.

Meta lists English, German, French, Italian, Portuguese, Spanish, Hindi and Thai as officially supported. It can often handle other languages, but quality drops. For Polish, Czech, the Nordic or Baltic languages, EuroLLM or Apertus are trained with much more European language data.

Change two things in your OpenAI client: the base URL, which points to the inference.eu API, and the API key. Then set the model string to \`llama-3.3-70b-instruct\`. Your messages, tools and streaming code stay the same.

No. Llama 3.3 70B is a text-only model. If you need to process screenshots, scanned documents or photos, use a multimodal model such as Gemma 4 31B, Qwen3.6 27B or Mistral Small 4, or DeepSeek-OCR 2 for documents.

Qwen3 235B A22B Instruct is a mixture-of-experts model with far more total parameters, a 256K context window and stronger results on coding and reasoning. Llama 3.3 is smaller, older and simpler, but very predictable. Many teams test both on their own prompts before deciding.

Yes. Llama 3.3 was trained to call tools, and on inference.eu you pass tools in the standard OpenAI \`tools\` parameter. The model returns tool calls in the usual format, so agent frameworks that work with OpenAI models work with Llama 3.3 too.

No. What you send to inference.eu is used only to produce your response. It is not used to train or fine-tune Llama 3.3 or any other model, and it is not shared with Meta. Meta only provides the weights; it is not involved in serving your requests.

You pay per token for the input and output you use, and every model sits on one invoice in euros. There is no separate contract per model. Get in touch with us if you expect large volumes and want to discuss terms.

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

## Related

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

-   [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.Chat](https://inference.eu/models/qwen3-235b-a22b-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.Chat](https://inference.eu/models/mistral-small-4-119b/)
-   [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.Vision](https://inference.eu/models/gemma-4-31b/)
-   [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