Pricing

Simple prices, drawn from a prepaid balance

Pay for GPU time, storage and tokens. No contracts, no monthly minimum. All prices in US dollars.

GPUs

Per GPU, per hour. Slices are a hard share of one card's memory and a guaranteed share of its compute; a whole card is yours alone.

GPU prices per hour
GPU	VRAM	On-demand	Per minute

Every GPU includes root access, SSH, a browser terminal, JupyterLab and a /workspace disk. Slices, spot and isolation

Exactly what the billing system does, in plain words.

Prepaid balance: Add credit by card, from US$25.00. GPUs, volumes and tokens draw from it. Auto top-up is optional.
Per minute: A running GPU is billed per minute at its hourly price divided by 60. Run it for 7 minutes and you pay for 7 minutes.
Launching: Launching sets aside 30 minutes of credit so a GPU can't run past an empty balance. Minutes you don't use stay in your balance.
Paused: A paused GPU keeps its slot and /workspace for 25% of its hourly price.
Terminated: Billing stops when you terminate.
Tokens: Inference is metered per token and debited from your balance every few minutes. Requests are refused once the balance is empty.

Included

Each GPU's own disk. Kept through pause and resume, deleted when you terminate.

US$0.10 / GB-month

Team disks that mount on several GPUs and outlive them. Charged by size while the volume exists, attached or not, metered per minute.

Questions about billing? The docs cover balances, top-ups and invoices.