from 24 GBChat
Private ChatGPT · Qwen3 32B
A chat in your browser for writing, code and documents. No subscriptions and no message limits.
from €0.30/h · RTX 3090, 24 GB
Run AI models · Llama
Run Llama 3 on a rented GPU
Llama 3.1 8B runs on any card with 12 GB or more. The 70B version needs about 50 GB in 4-bit: one data-centre card or several consumer ones. Llama 3.1 405B is for when answer quality matters most. Meta publishes all versions openly for assistants, chatbots and RAG.
- App
- PyTorch
- GPU
- RTX 3060, 12 GB
- Price
- from €0.12/h
To launch, your balance must cover one hour of the machine. You are billed by the minute for actual use.
from 2 GBof video memory
2–5 minutesuntil the server is up
By the minutebilled from your balance
SSH and Jupyterserver access
How much memory you need and which GPU to pick
Llama 3.2 3B takes about 2 GB of video memory in 4-bit; the cheapest suitable card is RTX 3060. Prices update from available catalogue machines when you open the page.
| Version | FP16 | 4-bit | Suitable GPUGPU | Price per hourPer hour |
|---|---|---|---|---|
| Llama 3.2 3B | 7 GB | 2 GB | RTX 3060 | €0.12 |
| Llama 3.1 8B | 18 GB | 6 GB | RTX 3060 | €0.12 |
| Llama 3.3 70B | 154 GB | 47 GB | 4× RTX 4060 Ti | €0.64 |
| Llama 3.1 405B | 892 GB | 268 GB | 8× L40 | €3.68 |
Estimate: FP16 ≈ 2 GB per billion parameters, 4-bit ≈ 0.6 GB, plus 10% for context. Pick: the cheapest configuration with enough memory; for large models that means several identical cards. All GPUs and prices
How to run Llama in 3 steps
- Sign up and top up. Create an account and top up from €10. To launch, your balance must cover at least one hour of the chosen machine; after that you pay by the minute.
- Choose a GPU. Memory for each version is in the table above, and the lightest one runs from 2 GB. The button on this page selects the cheapest option with the PyTorch template.
- Run the model. The server is ready in 2–5 minutes. Open Jupyter in your browser or connect over SSH and run the commands below.
Finished? Stop the server in the dashboard and billing stops. Save and stop keeps /workspace for €0.03 per GB per month, and it comes back on your next server automatically.
With Ollama (easiest)
curl -fsSL https://ollama.com/install.sh | sh
ollama run llama3.1:8b
With vLLM
pip install vllm
vllm serve meta-llama/Llama-3.1-8B-Instruct
What people run Llama for
01
An assistant for your team
A chat that can read internal documents through RAG. Llama 8B answers fast; 70B is more accurate on hard questions.
02
Fine-tuning for a task
QLoRA for 8B fits in 16–24 GB: teach the model your tone, company terms or an output format.
03
Processing text in bulk
Ticket classification, review summaries, dataset labelling. vLLM batches requests and keeps the card fully busy.
Tips before you launch
- For a web service with dozens of users, use vLLM: it batches requests and gets the most out of the card.
- Version 3.3 70B in 4-bit fits in 48 GB, and one RTX A6000 is cheaper than two 4090s.
- QLoRA on 16–24 GB is enough to fine-tune the 8B model.
Just need a chat? A ready-made app with no setup
If you need the model for conversations and don’t want to write code, launch Private ChatGPT: the server comes with Ollama, a Qwen3 model and the Open WebUI chat window in your browser.
from 12 GBChat
Private ChatGPT · Qwen3 8B
A fast chat on an affordable card: email drafts, translations, quick answers.
from €0.12/h · RTX 3060, 12 GB
Frequently asked questions
How much memory does Llama 3.1 8B need?
About 19 GB in FP16 and 6 GB in 4-bit, with room for context. An RTX 3090 or 4090 is comfortable.
Which GPU should I pick for the 70B version?
In 4-bit, 48 GB: an RTX A6000, an L40S or two 24 GB cards. In FP16, about 160 GB: 2× H100 or H200.
Do I need a separate licence?
The models are released under Meta’s licence: commercial use is allowed, with restrictions for very large services. Read the terms on the model page.
How long does a launch take?
The server is usually ready in 2–5 minutes, then SSH and Jupyter are available. Downloading the model depends on its size and is billed as normal server time.
What happens to my models and results after I stop?
When you stop, click Save and stop: the /workspace folder with your models, LoRAs and results goes to storage for €0.03 per GB per month and comes back on your next server automatically. If you stop without saving, the disk is deleted.
How is it billed?
By the minute from your balance in euros. Top up from €10; to launch, your balance must cover at least one hour of the chosen machine. Stop the server and billing stops.
A server is usually ready in 2–5 minutes
Sign up in a minute, top up from €10, pay by the minute. Stop the server and billing stops.