Llama 3 Hosting
Deploy Meta's Llama 3.3 models on dedicated European infrastructure.
Meta's Llama family is the workhorse of open LLMs: permissively licensed, well documented, and available in sizes from pocket models to 70B heavyweights. HostYourAI runs the Llama 3.1, 3.2 and 3.3 generations with vLLM on European GPUs, so you get Llama's quality without sending a single prompt outside the EU.
You can call Llama models in two ways. The shared EU Inference Router gives instant pay-per-token access, with popular sizes kept warm. For steady production traffic or strict isolation requirements, dedicated GPU instances run your chosen Llama with its own encrypted API key.
Why Llama hosting in the EU
Open weights, sovereign serving
Llama's weights are public, so nothing forces you to run them on American infrastructure. Hosting them on European GPUs turns an open model into a genuinely sovereign stack: the weights, the vLLM serving layer and the hardware are all outside any single vendor's control.
Data protection stays simple
Prompts to a Llama model here are processed in European datacenters and never used for training. With a DPA and a public subprocessor list, your GDPR documentation for the feature stays a page, not a project.
Pay for tokens, not for idle theory
On the Router you pay per token from a prepaid balance; a dedicated Llama instance bills per hour and idles down when unused. Both draw from one balance, explained on the pricing page, so experimenting with a bigger Llama is a decision you can reverse the same day.
EU Inference Router
‹ Insight 1 of 28 ›Choosing a Llama size for your workload
The Llama generations split neatly by job. The small 3.2 models (1B and 3B) are built for high-volume, low-stakes tasks: routing, tagging, extraction, guardrails. Llama 3.1 8B is the default choice for summarization and structured output, cheap enough to call on every request. At the top, Llama 3.3 70B delivers roughly the assistant quality people expect from a frontier chat model, with strong instruction following and multilingual output, including Dutch.
A practical pattern is to start every feature on 3.3 70B via the Router, measure quality, then push individual tasks down to smaller sizes until quality drops. The Model Garden shows which Llama variants are warm right now, and each can be tried in the Playground before you write code.
| Llama variant | Typical use |
|---|---|
| Llama 3.2 1B | Classification, routing, guardrail checks at very high volume |
| Llama 3.2 3B | Short summaries, entity extraction, autocomplete |
| Llama 3.1 8B | General summarization, structured JSON output, drafting |
| Llama 3.1 70B | Complex reasoning and long-context analysis |
| Llama 3.3 70B | Production chat assistants, multilingual dialogue, RAG answers |
from openai import OpenAI client = OpenAI( base_url="https://hostyourai.com/api/v1", api_key="hyai-...") client.chat.completions.create( model="llama-3.3-70b", messages=[{"role":"user","content":"Hallo!"}])
How to start
Llama models sit behind the standard OpenAI-compatible endpoint, so any OpenAI SDK works unchanged.
from openai import OpenAI
client = OpenAI(
base_url="https://hostyourai.com/api/v1",
api_key="hyai-...",
)
response = client.chat.completions.create(
model="llama-3.3-70b",
messages=[
{"role": "system", "content": "You are a concise Dutch-speaking assistant."},
{"role": "user", "content": "Leg in twee zinnen uit wat vLLM doet."},
],
max_tokens=200,
)
print(response.choices[0].message.content)Questions about Llama hosting
Which Llama versions are available?
The Llama 3.1, 3.2 and 3.3 generations, from 1B up to 70B. The Model Garden shows the live state of each: warm, warming up, serveable or on request.
Is Llama 3.3 70B kept warm on the Router?
Popular models are kept warm through a warm pool so requests start generating immediately. The catalog page for each model shows its current state before you call it.
Can Llama answer in Dutch?
Yes. Llama 3.1 and 3.3 are multilingual and handle Dutch well, which is one reason Dutch teams pick 70B for customer-facing chat.
How much GPU does a 70B Llama need?
A large card in the 80 GB class or a multi-GPU setup. You do not have to size this yourself: deployment enforces a VRAM floor, so a dedicated instance always gets hardware the model fits on.
Can I combine Llama with my own documents?
Yes. Link a knowledge base to the Router or to your instance and relevant context is injected into completions automatically. The guide Build RAG on EU GPUs shows the setup end to end.
Everything you need for AI
From model hosting to a customer-facing API, it is built for developers and businesses who want their AI running on infrastructure they actually control, inside the EU.
Your data and your models stay on European GPUs. GDPR-friendly by design.
Llama, Qwen, DeepSeek, Mistral, FLUX and plenty more. Pick one and it is warm in minutes, with no DevOps on your end.
Point your existing client at the Router and keep your tools. No rewrite, no lock-in.
From zero to a warm endpoint in minutes
No infra to manage. Pick a model, get an OpenAI-compatible URL, ship.
Pick a model
Choose from the Model Garden or paste any HuggingFace ID. Set the VRAM and pick an EU GPU.
Get your endpoint
We deploy vLLM, run readiness probes, and hand you a warm OpenAI- and Anthropic-compatible URL plus an API key.
Route and ship
Point your client at the Router. It auto-routes to a warm instance, idles GPUs when nobody is online, and logs every request.
Private by Default
HostYourAI keeps your models, prompts and data on European GPUs. It is built for teams that care about compliance, reliability and real control.
GPUs and data residency inside Europe. Your prompts never leave the EU.
Run open-weight models with no black boxes or hidden telemetry.
GPUs idle when nobody is online, so you only pay for what you run.
Your infra, your keys, your models. Leave whenever you want.
Works with the tools you already use
The Router speaks the OpenAI and Anthropic APIs, so it drops straight into the clients and SDKs your team already runs. Just change the base URL.
Try HostYourAI for free
Frequently asked questions
Can I run this in the EU?
Yes. HostYourAI runs open models on GPUs in European datacenters via vLLM. Your prompts and outputs never leave the EU and there is no US cloud provider in the chain.
Is it GDPR-compliant?
Yes. All processing happens inside the EU, a Data Processing Agreement (DPA) is available and the subprocessor list is public. Open weights also mean no training on your data.
Is the API OpenAI-compatible?
Yes. Point your existing OpenAI or Anthropic client at our Router (https://hostyourai.com/api/v1), change only the base URL and API key. No rewrite, no lock-in.
What does it cost?
Pay-as-you-go on one prepaid credit balance: the shared router per token or a dedicated GPU per hour. Free to start, no minimum, no fixed monthly fee.
Works with 390+ open models
Text and image models on dedicated EU GPUs. Every model tested on our own hardware.
Related pages
Explore more about EU-hosted AI on HostYourAI.
A ChatGPT alternative that runs on open models and European GPUs. OpenAI-compatible API, GDPR-friendly and without your data leaving the EU.
Read more →AI for healthcare on European GPUs. Patient data never leaves the EU, with open models, a Data Processing Agreement and GDPR-compliant processing.
Read more →Enterprise AI solutions for European companies. SLA guarantees, dedicated support, compliance certifications, custom integrations.
Read more →AI hosting that meets all GDPR requirements. Data stays in the EU, no US transfers, Data Processing Agreement included. For compliance-conscious companies.
Read more →Host StarCoder2 15B on dedicated NVIDIA A10 in European data centers. GDPR compliant, pay-as-you-go, OpenAI-compatible API.
Read more →European AI API with OpenAI compatibility. Easy migration, EU data centers, GDPR compliant, per-minute billing.
Read more →