Pricing

Pay-as-you-go, billed in USD. Model rates equal the provider's public price with zero markup; we take a 10% fee when you top up — that fee is our only revenue.

Why zero markup? Our value isn't in token arbitrage. It's in calling many providers through one integration, not juggling separate accounts and currencies, automatic fallback on failure, and one consolidated bill with per-request cost breakdowns you can audit. When a provider changes rates we follow — you always pay the public price.

Top up $20

$18.00 credited to your balance
  • Platform fee 10% ($2.00)
  • All models, no expiry
  • Per-request billing, accurate to 1e-9 USD
  • Requests return 402 when the balance runs out — you can never go into debt
Top up $20

Top up $50

$45.00 credited to your balance
  • Platform fee 10% ($5.00)
  • All models, no expiry
  • Per-request billing, accurate to 1e-9 USD
  • Requests return 402 when the balance runs out — you can never go into debt
Top up $50
Most common

Top up $100

$90.00 credited to your balance
  • Platform fee 10% ($10.00)
  • All models, no expiry
  • Per-request billing, accurate to 1e-9 USD
  • Requests return 402 when the balance runs out — you can never go into debt
Top up $100

Top up $500

$450.00 credited to your balance
  • Platform fee 10% ($50.00)
  • All models, no expiry
  • Per-request billing, accurate to 1e-9 USD
  • Requests return 402 when the balance runs out — you can never go into debt
Top up $500

Custom amounts supported (minimum $20). For larger commitments and enterprise rates, contact sales.

Rate card

Unit: USD per million tokens, identical to the provider's public rate card. Price table version 2026-09-11.1. For the full tables — including time-of-day and context tiers — see Models & pricing.

ModelInputOutputCache-hit inputPeak input / output
deepseek-flash$0.15$0.6$0.003$0.3 / $1.2
qwen-plus$0.1096$0.274$0.011—
qwen-max$0.3288$1.3151$0.0329—
qwen-flash$0.0205$0.2055$0.0021—
autoBilled at whichever model it routes to — routing itself is free

How billing works

🧾

Per-request pass-through

Each charge = tokens in that call × the provider's public rate. Expand any line in your bill to see the input / cache-read / output breakdown and the request_id.

⚡

Cache savings passed on in full

Input tokens that hit the provider's prompt cache are billed at the cache rate (as low as 1/50 of standard). Keep your system prompt stable and the savings are entirely yours.

🕐

Peak windows billed as-is

Where an upstream provider has peak/off-peak rates, we bill by the time the request actually happened. No markup, no subsidy.

🚫

Failed requests are free

Anything the gateway or the upstream returns an error for is never charged and never touches your balance.

🎁

Starter credit

Sign up and get $1 of credit — enough to run the whole integration end to end without adding a card.

🏢

Enterprise terms

Spending over $1,000/month? Talk to us about negotiated rates, dedicated concurrency, an SLA and invoicing.

Worked example A call using 1,853 input tokens (1,664 of them cache hits) and 5 output tokens on deepseek-flash, off-peak, costs 189 × $0.15/1M + 1,664 × $0.003/1M + 5 × $0.6/1M = $0.000036342. A $20 top-up ($18.00 credited) covers roughly 495,294 calls like that.