Pricing

Pay for the tokens you use. Nothing else.

Every model is billed per token at its listed price. You prepay a balance in crypto from $50, with no subscription, no seat fees and no expiry.

Create an account All model prices

The pricing model

Prepaid balance

Top up from $50 to $100,000 in USDT, USDC, ETH or BTC. The balance never expires.

Per-token prices

Each model lists a price per 1M input tokens and per 1M output tokens. That's the whole price list.

No plans or seats

No subscription, minimum spend, or per-key fee. Every model is available to every account.

Failures are free

A request that fails costs nothing; only completed responses are billed.

Current prices

The newest models from the major providers, straight from the live catalog. Prices are in USD per 1 million tokens; the model catalog lists every model.

Prices for featured models
Model	Input / 1M	Output / 1M	Model id
Anthropic Claude Haiku 5.5	$0.1	$0.5	`anthropic/claude-haiku-5.5`
Anthropic Claude Sonnet 5.5	$2.00	$10.00	`anthropic/claude-sonnet-5.5`
OpenAI GPT-6.1 Sol Pro	$2.00	$10.00	`openai/gpt-6.1-sol-pro`
OpenAI GPT-6.1 Sol	$2.00	$10.00	`openai/gpt-6.1-sol`
Google Gemini 4 Argon	$2.00	$10.00	`google/gemini-4-argon`
Google Nano Banana 2.1	$1.50	$7.50	`google/gemini-nano-banana-2.1`
SpaceXAI Grok 4.7	$2.00	$6.00	`x-ai/grok-4.7`
SpaceXAI Grok 4.6	$2.00	$6.00	`x-ai/grok-4.6`
DeepSeek DeepSeek V4.1 Flash	$0.02	$1.00	`deepseek/deepseek-v4.1-flash`
DeepSeek DeepSeek V4 Pro 0423	$0.9553	$1.91	`deepseek/deepseek-v4-pro`
Qwen Qwen3.8 Max Prime	$4.00	$12.00	`qwen/qwen3.8-max-prime`
Qwen Qwen3.8 Omni Flash	$0.15	$0.47	`qwen/qwen3.8-omni-flash`
MoonshotAI Kimi K3.1 Pro	$2.00	$10.00	`moonshotai/kimi-k3.1-pro`
MoonshotAI Kimi K3	$0.161	$20.00	`moonshotai/kimi-k3`
Mistral Mistral Large 4	$1.36	$4.18	`mistralai/mistral-large-4-0`
Mistral Mistral Medium 3.5	$1.50	$7.50	`mistralai/mistral-medium-3-5`

Prices follow what providers charge and are updated by a daily catalog sync. A price change applies to requests made after it.

Cost calculator

Estimate a workload: pick a model, set typical token counts and volume. It uses the same formula as billing. A token is roughly ¾ of an English word, so 1,000 tokens is about 750 words.

ModelInput tokens per requestOutput tokens per request

Requests per day

Per request: $0.00045
Per day: $0.45
Per 30 days: $13.50

How a charge works

Per-request cost

cost = input_tokens  × input_price  / 1,000,000
     + output_tokens × output_price / 1,000,000

# Example: 2,000 input + 500 output tokens at $3 / $15 per 1M
#   2,000 × 3  / 1,000,000 = $0.0060
#     500 × 15 / 1,000,000 = $0.0075
#   total                  = $0.0135

Token counts come from the provider and are returned in every response's usage object, so you can check each charge.
Long-prompt tiers. Some providers charge more above a prompt-length threshold. When the provider reports a higher cost than the listed price gives, that cost is charged.
Exact arithmetic. Costs are computed in decimal, not floating point, and every request's cost is shown in your logs to six decimal places.
Streams you cancel are billed for the tokens generated before you disconnected.

Why a request can be refused with money in the account

Before a request runs, its worst-case cost (your input plus the maximum output allowed) is reserved from your balance. Setting max_tokens to what you need keeps that reservation small. See cost reservation.

Topping up

1
Choose an amount
Anything from $50 to $100,000.
2
Pick a coin
USDT, USDC, ETH or BTC. The exact coin amount is set by the exchange rate when the payment is created.
3
Send the exact amount
Scan the QR code or copy the address, and send within the time shown on the payment screen.
4
Get credited
Your balance updates when the payment is confirmed on-chain, and an invoice appears under Invoices.

Send the exact amount, on the right network

Sending a different coin, using a different network, or underpaying can leave a payment unconfirmed. Check the coin and amount on the payment screen before you send.

Spending controls

Controls that limit spend
Control	What it does
Prepaid balance	Requests are refused rather than taking your balance below what they could cost.
Per-key monthly cap	Each key can have its own monthly limit; requests over it are refused with 402.
Per-key model allowlist	Restrict a key to the models a workload needs, so a typo can't hit an expensive model.
Low-balance email	An email when your balance falls below $5, at most once a day.
Request logs	Every request's model, tokens and cost, filterable and exportable as CSV.

FAQ

Is there a free tier?

No. The catalog only lists models with a per-token price, and accounts start with a top-up from $50. There's no subscription to cancel, and unused balance stays yours.

Does my balance expire?

No. Balance never expires, whether you use it this week or next year.

Do failed requests cost anything?

No. Only completed responses are billed. Upstream errors, refused requests and timeouts before a response cost $0.

Can I get an invoice?

Yes. Every confirmed top-up creates an invoice you can view and print from the Invoices page.

Can I pay by card?

Not at the moment. Top-ups are crypto only: USDT, USDC, ETH or BTC.

Model catalog

Every model with its price and context window.

How routing works

Reservation, settlement and what's retried.