On this page
A curated model catalog
The Model Garden lists serveable open models for chat, image and embeddings. For each model you see its state: warm, EU and whether it is warming up. So you always know what is ready to run right now.
From catalog to endpoint
Pick a model from the Garden or paste any HuggingFace ID. You deploy it on a European GPU through vLLM and get an OpenAI- and Anthropic-compatible URL plus an API key. Want to test first? With the Playground and the "Verify it works" step you check a model before you commit to it.
Chat, image and embedding models
- Text models such as Llama, Qwen, Mistral, DeepSeek and Gemma
- Image models such as FLUX and SDXL
- Embedding models for RAG and search
- Live warm/EU/warming-up state per model
How to deploy a model
Pick a model, set the VRAM and deploy on a European GPU. You get an OpenAI-compatible endpoint you can call right away:
from openai import OpenAI
client = OpenAI(
base_url="https://hostyourai.com/api/v1",
api_key="hyai-...",
)
resp = client.chat.completions.create(
model="qwen3-8b",
messages=[{"role": "user", "content": "Hello"}],
)
print(resp.choices[0].message.content)Want to deploy a model step by step yourself? Read the guide Deploy your own LLM with vLLM.
What the states mean
| State | Meaning | What you do |
|---|---|---|
| Warm | The model is running and answers instantly | Call it right away through the Router |
| Warming up | The model is being started right now | Wait a moment; the state refreshes live |
| Serveable | Tested and ready to start | Call or deploy it; the first start takes a moment |
| On request | Available but not yet verified by us | Ask via contact and we set it up |
Verified means we actually ran and checked the model on our own EU hardware, not a paper promise.
Questions about the Model Garden
How many models are in the catalog?
Over a hundred open models for chat, code, image and embeddings. The live list with pricing is on the public catalog.
My model is not listed?
Paste a HuggingFace ID and deploy it yourself on a dedicated instance. Your own or fine-tuned models work the same way.
What does "verified" mean?
That we ran and tested the model ourselves on our EU GPUs. Serveable-but-unverified models are marked "on request", so you never face surprises.
How do I test a model before integrating it?
Open it in the Playground and chat with it directly. From the Garden you can preselect a model in one click.
Are there image and embedding models too?
Yes. Image models such as FLUX and SDXL and embedding models for RAG and search, next to the chat and code models.
What does a Garden model cost?
Through the Router you pay per token, per model class. On a dedicated GPU you pay per hour. Both from one prepaid balance; see the pricing page.
