Credits & Billing

ParalonCloud runs on a single credit balance. You top it up once, and the same balance pays for everything on the network — GPU rentals and inference alike. There's no subscription and no monthly minimum; you pay only for what you use.

Topping up

Add credits on the Add Credits page by paying with crypto — USDC or USDT on Ethereum or Solana, by connecting your wallet.

The minimum top-up is $5, credits are added instantly, and they never expire.

How rentals are billed

GPU rentals are pay as you go, billed per minute:

• A short 1-hour balance is required up front to open a session — this reserves room so a rental can't run your balance negative.
• You're charged only for the minutes you actually use, not the reserved hour.
• Each node's hourly rate is either the platform price for its GPU or a custom price set by its provider — whatever is shown on the Rent page is what you pay.
• The rate is locked when your rental starts — price changes, by the provider or the platform, never affect a running session.
• Stop anytime from My Rentals. Billing ends when the rental ends.

How a rental splits

Every rental price is divided between the provider and the platform:

Party	Share of each rental
Provider	80% — always
Platform fee	14%–20%
Referrers	up to 6% (15% of the platform fee per referred side)

A few things worth being clear about:

• Providers always keep 80%. Your share for hosting never changes.
• The platform fee is up to 20%, and it's the only part referrals touch.
• Referral rewards come out of the platform fee, not the provider's 80%. When a rental involves a referred renter or provider, that referrer earns 15% of the platform fee for their side — about 3% of the rental. So when both the renter and the provider were referred, the platform's effective fee drops to as low as 14%, while the provider still keeps their full 80%.

How inference is billed

Inference is billed per token, from the same credit balance, at the list price of the model you call. What you pay depends on the type of API key the request came in on. There are three.

Key type	Who it is for	What it costs	Limits
Free	trying the API	nothing — a 100,000-token trial, counted across every request the key makes, for the life of the key, plus 5 images a day	one free key per account, 5 requests/min, 2 images/min
Premium	building on the API	per token and per image, at list price, from your credits	as many keys as you want, 60 requests/min, 300 images a day
Partner	organisations with their own agreement	metered, invoiced outside the platform	set up per email domain — write to us

The free trial

Every account can hold one free key. It carries a trial of 100,000 tokens — input and output added together — and every request draws from it. The Console shows how much of the trial is left. When it is used up, requests on that key are refused with a 402 (free_quota_exhausted) until you create a premium key. The free key cannot be revoked; if you lose it, regenerate it from the Console and the trial balance stays.

Images are part of the trial too: a free key can render 5 a day, at most 2 a minute, and they are not charged.

Premium keys

A premium key bills each request the moment it completes: tokens in × input price + tokens out × output price, to eight decimals, deducted from your credits. Images are priced per image. You can create as many premium keys as you need, name them per project, and see what each one has cost under Usage in the Console.

A request is refused up front, with a 402, if your balance cannot cover it. Nothing is ever charged for a request that failed.

Prices

Prices are per model and are shown in the Console under Available Models before you pick one. Today, per million tokens:

Model	Input	Output
qwen3.8-27b	$0.12	$1.70	vision, tool calling, 32k context
qwen3-3b	$0.05	$0.20	16k context
gemma3-4b	$0.04	$0.09	8k context
z-image-turbo	—	—	$0.003 per image

Every chat completion response includes a usage block (prompt_tokens, completion_tokens, total_tokens), so a request's cost can be worked out from the response alone.

Where the money goes

The split is the same as for rentals: 80% of every request goes to the provider whose GPU served it, credited the moment the request completes, and the platform keeps 20%. Requests on free keys draw on the trial, not on credits, so they carry no split. The mechanics are described in Every Inference Request Pays the GPU That Ran It.

Watching your balance

• Rentals — see live consumption and remaining balance under My Rentals.
• Inference — track requests, tokens and credits spent per API key and per model under Usage in the Console; free keys show the trial remaining.
• History — every top-up and charge is listed in your Transactions. Inference is shown as one line per day, with the request count, tokens and models; the per-request records sit underneath and are what the balance is computed from.

Running low mid-session? Top up on Add Credits — credits land instantly, so you don't have to stop your work.