Quickstart
Set up your account, install the SDK, and launch your first GPU instance in minutes.
Prerequisites
- A valid email address
- Python 3.8+ (for SDK) or any HTTP client
- Funds for GPU rental (card or SBP via T-Bank, or card via Stripe)
1. Create Your Account
Sign Up
Visit gpuniq.com and click Sign Up. Use email or Google Sign-In.
Verify Email
Enter the 6-digit verification code sent to your inbox.
Add Funds
Go to Balance in your dashboard and deposit by card or SBP through T-Bank, or by card through Stripe.
2. Get Your API Key
Go to LLM API Keys in your dashboard and create a new key. Your key starts with gpuniq_ — save it securely.
API keys work across all endpoints: marketplace, instances, volumes, LLM, and the OpenAI-compatible proxy at /v1/openai (for Claude Code, Cursor, LiteLLM).
3. Install the SDK
pip install GPUniq
4. Your First GPU Rental
Option A: Dex-Cloud (Simplest)
Pick a GPU type and deploy — the platform finds the best machine automatically.
from gpuniq import GPUniq
client = GPUniq(api_key="gpuniq_your_key")
# See available GPU types
gpus = client.gpu_cloud.list_instances()
for gpu in gpus["featured"]:
print(f"{gpu['gpu_name']}: ${gpu['gpu_price_per_gpu_hour_usd']}/hr")
# Deploy an RTX 4090
deploy = client.gpu_cloud.deploy(
gpu_name="RTX_4090",
docker_image="pytorch/pytorch:latest",
disk_gb=100,
)
print(f"Deploying... Job ID: {deploy['job_id']}")
Option B: Marketplace (Full Control)
Browse all available servers and pick a specific machine.
# Browse GPUs with filters
gpus = client.marketplace.list(
gpu_model=["RTX 4090"],
min_vram_gb=24,
sort_by="price-low",
)
print(f"Found {gpus['total_count']} GPUs")
# Rent a specific machine
agent_id = gpus["agents"][0]["id"]
order = client.marketplace.create_order(
agent_id=agent_id,
pricing_type="hour",
docker_image="pytorch/pytorch:latest",
)
5. Connect to Your Instance
After deployment, find SSH credentials in your dashboard or via API:
instances = client.instances.list()
for inst in instances["instances"]:
print(f"Task {inst['task_id']}: {inst['status']}")
ssh root@<host> -p <port>
# Password shown in dashboard
6. Verify GPU Access
nvidia-smi
python -c "import torch; print(torch.cuda.is_available())"
7. Set Up Command Checkpointing (Optional)
Use gg to checkpoint your commands — if your instance restarts, replay them instantly:
# Initialize gg with your instance token (shown in dashboard)
gg init <your-gg-token>
# Run commands with checkpointing
gg run pip install torch transformers
gg run python train.py --epochs 100
# After a restart, replay unfinished commands
gg replay
See the CLI client and CLI on a GPU instance pages for full details.