FAQ
Do I need to verify my identity?
Section titled “Do I need to verify my identity?”No. You sign in with your email and a one-time code. No documents are required — neither for renting nor for hosting.
Which payment methods are supported?
Section titled “Which payment methods are supported?”USDT on the TRON network (TRC20). See Add funds with USDT.
I sent funds on the wrong network. Can I get them back?
Section titled “I sent funds on the wrong network. Can I get them back?”Unfortunately not. Only USDT on TRON (TRC20) is credited; other tokens and networks cannot be recovered.
What happens to my unused balance?
Section titled “What happens to my unused balance?”It stays on your account and never expires.
How am I billed?
Section titled “How am I billed?”Per second, only while a server is running. Downloading the image and the model is free, and nothing is charged while a host machine is offline. See GPUs and pricing.
What if my balance runs out?
Section titled “What if my balance runs out?”The server is stopped (not deleted). Its data is kept for 7 days; add funds and start it again.
Will my data survive a stop?
Section titled “Will my data survive a stop?”Yes, /workspace stays on the host machine while the server is stopped — for up to 7 days. Destroying the server deletes it. Keep your own copies of important results.
Can I start a stopped server on another machine?
Section titled “Can I start a stopped server on another machine?”Not directly: the data lives on the original machine. If its GPUs are busy, copy your data out (or keep it in your own storage) and launch a new server.
Can I combine GPUs from different servers?
Section titled “Can I combine GPUs from different servers?”GPUs are combined within one machine — up to 8 per server, used together by vLLM/SGLang or distributed training. See Multi-GPU.
Can I use my own Docker image?
Section titled “Can I use my own Docker image?”Yes, with the Custom Docker image template (public images). See Templates.
How do I call my model from code?
Section titled “How do I call my model from code?”Create an API key and use the proxy URL, e.g. with the OpenAI SDK. See API.
Why does the first launch of a model take longer?
Section titled “Why does the first launch of a model take longer?”The first launch downloads the weights from Hugging Face. We cache them, so later launches (by anyone) start much faster. See Models.
Can I rent out my own GPU?
Section titled “Can I rent out my own GPU?”Yes. Install one command on a Linux machine with a supported NVIDIA GPU. See Rent out your GPU.
How does the referral program work?
Section titled “How does the referral program work?”You get 5% of what your invited users spend on GPUs for 12 months. See Referral program.
How is my account protected?
Section titled “How is my account protected?”With one-time email codes, server-side sessions and hashed API keys. See Security.