Pricing and billing

How billing works

Accounts are prepaid. Credits are purchased in USD and consumed per request at the per-token prices shown in the console. No subscription is required; BYOK plans are optional.

Purchased credits do not expire while the account remains active. Promotional or trial credits may carry an expiry date, which is shown when they are granted.

Accounts that sign up in the console and verify their email receive $1 in trial credit, once per account. Trial credit is valid for 7 days, is offered while this month's trial budget lasts, and unused trial credit expires after the 7 days. After that, top up or use your own keys with the free BYOK plan.

Token prices

Every text model's input, output and cache-read price is published on the Models page.

Where a model's route supports prompt caching, cached input bills at the cache-read price published for that model — at or below the provider's own cache-read price, DeepSeek V4 Pro, DeepSeek V4.1 Flash and Kimi K3 excepted — and cache writes carry no surcharge. The console shows the live price applied to each request, which prevails over any page.

See all model prices →

Image and video

Image generation is billed per image, listed separately from text and video models. Call POST /v1/images/generations; images come back in the response as base64 (b64_json). Failed requests are not charged.

Video generation is billed per second of output (reference-image inputs per image; Seedance 2.5 and 2.0 per video token, settled on actual output), listed separately from text models. Submit via POST /v1/videos, poll the task, download through platform URLs. Failed or cancelled tasks are not charged.

Top-ups, fees and taxes

Card via Stripe, USDT, or Wise bank transfer in USD. Every top-up gets an invoice automatically, with the header you set in the console.

Where applicable, tax is calculated by Stripe Tax at checkout based on the billing address, shown before payment, and recorded on the order. Card top-ups carry a payment processing fee (5.5%, or $0.80 minimum, shown before payment); credited amounts always equal the principal. USDT top-ups carry a 1% fee ($0.10 minimum). The minimum card top-up is $5. Business customers can add tax IDs.

When the balance runs out

Requests are rejected without creating a negative balance. When the balance can't cover a full-length response and the request doesn't set max_tokens, the output limit is lowered to what the balance covers and the response ends with finish_reason "length". Set a balance alert in the console to get warned early.

Interrupted streaming requests

If a streaming request is closed before it completes, it is billed for the usage the provider reports for that request. This can include reasoning tokens and output the provider finishes generating shortly after the connection closes. If the provider does not report usage in time, usage is estimated: for a model that produced reasoning, as the input plus the full output allowance reserved for the request (max_tokens, or the default allowance when none is set); otherwise, from the text generated. Estimates never exceed the amount reserved for the request. If the provider fails mid-stream, the request is not charged.

Bring your own keys

BYOK lets you store your own provider keys so inference runs on your provider accounts and contract prices, with the platform as automatic fallback. The free plan includes 1,000,000 BYOK requests per month with up to 3 providers, and needs no top-up. See the BYOK page.

Requests switch to the platform's paid lines automatically at the published platform prices — no hidden markup — and your application keeps running. The monthly quota resets on the 1st of the month (UTC); models that run only on your own key pause until then.

BYOK plans →