Data centre in Poland · VAT invoice · support in Polish

GPU power for AI.
By the hour, by the month or via an API.

NVIDIA RTX 4000 Ada with 20 GB of memory for training, fine-tuning and running models. You only pay for what you use, and your data stays in Poland.

View prices Order a subscription

20 GBGDDR6 ECC VRAM

from 0,16 €/hnet hourly rate

~1,000 tokens/svia the API when there are multiple queries

PLserver room and data in Poland

Available cards

Two configurations in our server centre in Poland. A clear price, with no bidding and no hidden transfer fees.

available immediately

NVIDIA RTX 4000 Ada Generation

GPU memory20 GB GDDR6 with ECC

ArchitectureAda Lovelace, 6,144 CUDA cores

Tensor cores4th generation, FP8 / FP16 / BF16

EnvironmentLinux, CUDA, PyTorch, JupyterLab

AccessSSH, JupyterLab or an OpenAI-compatible API

LocationPoland, Lexeme data centre

0,46 € / net hours

You pay for the time the machine is switched on; 0,27 € per hour with the pay-as-you-go package, or 134,35 € per month with the subscription.

Set it for a specific time Subscription 134,35 € / month

4 cards, from the hand

NVIDIA GeForce RTX 3080

GPU memory10 GB of GDDR6X per card

ArchitectureAmpere, 8,704 CUDA cores

In a single machine1 to 4 graphics cards (up to 40 GB VRAM)

On the card1 CPU core, 3 GB RAM

AccessSSH and JupyterLab, PyTorch with CUDA

LocationPoland, Lexeme data centre

0,16 € / hours per card

You pay for the time the machine is switched on, with a monthly bill; 4 cards = 0,66 €/h. Permanently 24/7 with no hour limit: 68,08 € net per GPU per month.

Set it for a specific time Subscription 68,08 € / month

Three methods of settlement

Choose the model that suits your work. Net prices; a VAT invoice will be issued for each payment.

By the hour

from 0,16 € / hours

RTX 3080: 0,16 €/h per card, 1–4 cards per machine
RTX 4000 Ada: 0,46 €/h or a 50-hour package for 13,66 € / 200 hours for 54,65 € (0,27 €/h)
You only pay for the time the machine is switched on; you’re billed once a month
You can enable and disable the security spending limit in the control panel

RTX 3080 by the hour RTX 4000 Ada by the hour RTX 4000 Ada Hourly Package

Best suited to continuous operation

Monthly subscription

134,35 € / month

A card just for you, 24 hours a day
Approx. 0,19 € per hour for continuous work
Full SSH access, your own environment
Guaranteed availability, without interruption

Model APIs (tokens)

0,46 € / 1 million tokens

A pack of 10 million tokens for 4,55 € or 50 million for 22,77 €
OpenAI-compatible APIs, open-source models
No server management
The data does not leave Poland

Buy a token pack

How much does it actually cost?

Example monthly costs for a single RTX 4000 Ada graphics card (net).

Directions for use	Hours per month	By the hour	Subscription
Training sessions after work, 2 hours a day	approx. 44 hours	approx. 12,07 €	134,35 €
Working day: 8 hours for 22 days	176 h	approx. 48,05 €	134,35 €
Working day and night, 16 hours a day	approx. 480 hours	approx. 131,16 €	134,35 €
All the time, 24/7	approx. 720 hours	approx. 196,74 €	134,35 €

If you use it for more than around 490 hours a month, a subscription works out cheaper.

How it works

From placing an order to working on the card in just a few minutes.

Select and pay

A package of hours, a subscription or tokens. You pay online and receive your invoice straight away.

Get your login details

The address, SSH login or API key can be found in the customer portal under the ‘My Services’ tab.

Run the model

Ollama, vLLM, PyTorch or your own Docker container. We’ll help you get started.

Ready-made tool sets

The card comes with the software environment pre-installed – so you don’t waste time setting up drivers and libraries.

OllamaLocal language models with a single command

vLLMFast model serving via API

PyTorch + CUDAA ready-to-use environment for training and fine-tuning

Jupyter NotebookWorks in your browser without any configuration

Stable Diffusion / ComfyUIGenerating images and graphics

WhisperSpeech-to-text transcription

Open WebUIA ChatGPT-style chat using your own models

DockerYour own container, full control

What is it used for?

Chatbots and virtual assistantsLanguage models 7B–14B trained on our own data

Model tuningLoRA and QLoRA on 20 GB of VRAM

Transcript of the speechWhisper, film subtitles

Graphics and videoStable Diffusion, image generation

Document analysisOCR, computer vision models, invoices

Batch processingProduct descriptions, translations, classification

RenderingBlender, CUDA

Research and prototypesJupyter, PyTorch, TensorFlow

OpenAI-compliant API

With the token package, you simply change the address and key in your application. We support open-source models running on our server in Poland – you can integrate them into your own code, an AI-powered editor, or a popular chat app that allows you to specify your own API provider.

Buy a token pack

What is included in the price of 0,46 € per 1M tokens?

Your data and enquiries are never used to train models – neither ours nor anyone else’s
A token dedicated exclusively to queries within the package; you do not share it with other customers, as is the case with cheap, mass-market APIs
No limits on the number of queries and no throttling, as is common with large, public providers
Data and enquiries do not leave the Polish data centre

cURL

curl https://api.cyberplaneta.pl/v1/chat/completions \ -H "Authorisation: Bearer $CYBERPLANETA_KEY" \ -H "Content-Type: application/json" \ -d '{"model": "qwen2.5-7b", "messages": [{"role": "user", "content": "Hi!"}]}'

Python (OpenAI library)

from openai import OpenAI client = OpenAI( base_url="https://api.cyberplaneta.pl/v1", api_key="YOUR_CYBERPLANETA_KEY", ) resp = client.chat.completions.create( model="qwen2.5-7b", messages=[{"role": "user", "content": "Hi!"}], ) print(resp.choices[0].message.content)

Cursor

Settings → Models → enable ‘Override OpenAI Base URL’, paste the URL and key, and add the model name as a custom model.

Base URL: https://api.cyberplaneta.pl/v1 API key: YOUR_CYBERPLANETA_KEY Model: qwen2.5-7b

Open WebUI, Chatbox and similar chat applications

In the provider settings, select ‘Custom API address / OpenAI-compatible’ and enter the same address, key and model name as above – that’s it, you can now chat with our card just like you would with ChatGPT.

Start from 13,66 €

A 50-hour package on the RTX 4000 Ada. No fixed-term contract.

Buy a package of hours

GPU power for AI.By the hour, by the month or via an API.